| openstackgerrit | Jim Rollenhagen proposed openstack/ironic-python-agent: Support configdrive from HTTP(S) https://review.openstack.org/135792 | 00:10 |
|---|---|---|
| jroll | ^ beginning of config drive work, working bottom-up here | 00:10 |
| *** imtiaz has joined #openstack-ironic | 00:13 | |
| *** ryanpetrello has joined #openstack-ironic | 00:13 | |
| *** marcoemorais has quit IRC | 00:14 | |
| openstackgerrit | Ruby Loo proposed openstack/python-ironicclient: Add 'API' to description of ironic command https://review.openstack.org/135794 | 00:24 |
| jroll | rloo: hey, can you remove your -2 here? :) https://review.openstack.org/#/c/128388/ | 00:31 |
| rloo | sure, sec | 00:31 |
| jroll | thanks | 00:31 |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic: Support configdrive in agent driver https://review.openstack.org/128388 | 00:32 |
| *** naohirot has joined #openstack-ironic | 00:32 | |
| naohirot | good morning ironic | 00:33 |
| jroll | heya naohirot | 00:33 |
| rloo | jroll: you're good to go. But you might want to ask devananda to approve that spec ;) | 00:33 |
| naohirot | jroll: hi thanks for yesterday. | 00:34 |
| devananda | morning, naohirot ! | 00:34 |
| devananda | rloo: which spec? | 00:34 |
| naohirot | rloo: good morning | 00:34 |
| devananda | oh, nm | 00:34 |
| naohirot | devananda: good morning | 00:34 |
| rloo | https://blueprints.launchpad.net/ironic/+spec/expose-configdrive | 00:34 |
| rloo | morning naohirot | 00:35 |
| jroll | devananda: I feel like we should make you a button that +A's a spec and approves the associated blueprint | 00:35 |
| * NobodyCam is back | 00:35 | |
| jroll | or I can jump onto storyboard instead in an attempt to kill launchpad faster | 00:35 |
| *** ryanpetrello has quit IRC | 00:38 | |
| devananda | jroll: I have one | 00:41 |
| devananda | I just haven' tused it today | 00:41 |
| jroll | ha | 00:42 |
| jroll | devananda: any opinions on how configdrive in the iscsi driver works? | 00:42 |
| devananda | eeeeh | 00:43 |
| devananda | nope | 00:43 |
| jroll | ok | 00:43 |
| JayF | how is it that hard? just a second image to plop into a partition | 00:43 |
| devananda | it doesn't fit within that driver's space at all -- unless one uses virtualmedia | 00:43 |
| devananda | what partition? | 00:43 |
| JayF | devananda: sure it does | 00:43 |
| JayF | devananda: you create one | 00:43 |
| JayF | that's what we do in IPA today | 00:43 |
| devananda | hm | 00:43 |
| jroll | JayF: however, most users of iscsi today use root/ephemeral/swap, which add up to total disk size | 00:44 |
| jroll | (I think) | 00:44 |
| devananda | yes | 00:44 |
| JayF | Can we steal 64mb from the root? | 00:44 |
| JayF | heh | 00:44 |
| jroll | so even carving out 64mb might be problematic | 00:44 |
| jroll | it depends | 00:44 |
| jroll | on flavor configuration and user options | 00:44 |
| devananda | you'll break some assumptions of the driver // of existing deployers | 00:44 |
| jroll | yeah | 00:44 |
| jroll | though | 00:44 |
| JayF | only if they turn on | 00:44 |
| devananda | so backwarrsd compat is an issue | 00:44 |
| JayF | a non default option | 00:44 |
| devananda | sure | 00:44 |
| jroll | this will be disabled by default | 00:44 |
| devananda | that's reasonable | 00:44 |
| jroll | ok, I think I'll jfdi | 00:45 |
| devananda | :) | 00:45 |
| jroll | going to put it before root partition though | 00:45 |
| JayF | jroll: -1 | 00:45 |
| jroll | so I don't have to deal with calculations | 00:45 |
| devananda | nope | 00:45 |
| jroll | why | 00:45 |
| JayF | jroll: that's going to break some bootloaders | 00:45 |
| devananda | that | 00:45 |
| JayF | jroll: and older hardware, etc etc | 00:45 |
| devananda | and, look at the code already doing calculations. they are non-trivial today | 00:45 |
| devananda | you do that, you break them | 00:45 |
| jroll | sigh | 00:45 |
| jroll | I'm not entirely sure about that | 00:46 |
| devananda | jroll: ironic/common/disk_partitioner.py | 00:46 |
| NobodyCam | jroll: how about somehting like: sudo mount -t tmpfs -o size=64M tmpfs /tmp/config_drive | 00:46 |
| NobodyCam | :-p | 00:46 |
| JayF | You can't use a tmpfs for configdrive | 00:46 |
| JayF | even if you were inside the image and could make it exist | 00:46 |
| JayF | cloud-init won't look there | 00:46 |
| jroll | devananda: doesn't look like complex math | 00:47 |
| devananda | jroll: and drivers/modules/deploy_utils.py is where the calculations are actually done | 00:47 |
| jroll | right | 00:47 |
| jroll | there's no complex math afaict | 00:48 |
| devananda | I didn't say "math" :) | 00:48 |
| jroll | deploy_utils just calls add_partition in a given order | 00:48 |
| jroll | math/calculations | 00:48 |
| jroll | though idk enough about all this | 00:49 |
| devananda | jroll: destroy_disk_metadata | 00:49 |
| jroll | it goes ephemeral/swap/root right now, is that what bootloaders assume? | 00:50 |
| devananda | there is no local bootloader | 00:50 |
| devananda | today :) | 00:50 |
| jroll | right | 00:51 |
| jroll | so | 00:51 |
| jroll | I don't see any assumptions in ironic's code that will break | 00:51 |
| jroll | idk of any other assumptions that will break | 00:51 |
| devananda | jroll: where does cloud-init look for the config drive? | 00:52 |
| jroll | on the partition with label config-2 | 00:53 |
| devananda | ok, so by label | 00:53 |
| jroll | so I'm indifferent as to where to put it | 00:53 |
| jroll | I just don't want to put code to do (root - 64mb) if possible | 00:53 |
| jroll | though that's shorter than this conversation :P | 00:53 |
| devananda | heh | 00:53 |
| devananda | root will autogrow | 00:54 |
| devananda | swap and ephemeral will not | 00:54 |
| devananda | that's why root is at the end, IIRC | 00:54 |
| jroll | right | 00:54 |
| JayF | devananda: by label ... and some other stuff | 00:54 |
| devananda | so configdrive, ephemeral, swap, root seems fine to me | 00:54 |
| jroll | what we do in onmetal, is put it on the last 64mb | 00:54 |
| JayF | devananda: aka by label, but filtered | 00:54 |
| JayF | devananda: we already have a patch upstream to stop filtering out individual partitions :) | 00:54 |
| jroll | JayF: will first partition work? | 00:55 |
| JayF | yes; again discounting that that will lolbreak preexisting images that use hardcoded devices in fstab and will make it difficult to get a bootloader on those systems | 00:56 |
| jroll | meh, if people start using this, they'll presumably have to change their images anyway yeah? | 00:57 |
| devananda | configdrive data can change during a nova rebuild, yes? | 00:57 |
| devananda | jroll: wait. why would they have to change the image? | 00:57 |
| jroll | devananda: to add cloud-init? | 00:58 |
| devananda | assume cloud-init is there, and using the metadata service today | 00:58 |
| jroll | dunno, they may already have that, fell free to ignore that comment | 00:58 |
| devananda | jroll: are you thinking that the use of configdrive is a deployer config option, or a per-node at-deploy-time decision? | 00:59 |
| devananda | *with the iscsi deploy | 00:59 |
| devananda | (i hope you say the former) | 00:59 |
| jroll | it can be both | 00:59 |
| devananda | eh, no | 00:59 |
| devananda | because the partition table of a node can not be changed during a rebuild | 00:59 |
| JayF | deployer config+++ | 00:59 |
| jroll | https://github.com/openstack/nova/blob/master/nova/virt/configdrive.py#L179-193 | 00:59 |
| jroll | if you have the configdrive on the last 64mb, you could certainly add/remove that on rebuild | 01:00 |
| jroll | I would think | 01:00 |
| JayF | That's interesting | 01:00 |
| devananda | so a node which was deployed without a local configdrive partition can not be rebuilt (via nova rebuild) with a cnofigdrive | 01:00 |
| JayF | personally I'd almost want to treat the configdrive partition like I would in preserve-ephemeral | 01:00 |
| devananda | jroll: no. because root will autogrowfs to the full disk size | 01:00 |
| JayF | like leave the partition there; change the contents | 01:00 |
| devananda | JayF: yes | 01:00 |
| devananda | if using local configdrive, always write the partition, even if it's empty | 01:00 |
| JayF | devananda: cloud-init autogrowfs works great with the configdrive at the end of the drive fwiw :) | 01:01 |
| devananda | JayF: I mean, if there WASN"T a configdrive when the node was first deployed, you can't "nova rebuild" to add one at the end of the disk | 01:01 |
| JayF | I agree | 01:01 |
| JayF | devananda: well, I was more saying; on nova rebuild --preserve-epehemeral, I'd just put new content in the preexisting configdrive | 01:01 |
| devananda | because shrinkage | 01:01 |
| devananda | JayF: ah. yes. totally. | 01:01 |
| devananda | but the part table isn't changing | 01:02 |
| JayF | exactly | 01:02 |
| JayF | so just like IPA does today | 01:02 |
| devananda | so config option for the driver | 01:02 |
| JayF | whatever iscsi does should prefer to write to a preexisting, labelled partition | 01:02 |
| JayF | that way even when you get to the point of iscsi doing full disk images like IPA does | 01:02 |
| devananda | and when using local config drive, always make the partition | 01:02 |
| devananda | even if empty | 01:02 |
| JayF | the logic matches: a full disk image with "config-2" labelled partition anywhere gets the configdrive written there; if it doesn't exist it's added | 01:02 |
| devananda | JayF: ooh. gotcha. you'll put it at the end, then. since you can't write a whole-disk image and then stick a partition at the beginning | 01:03 |
| JayF | exactly | 01:04 |
| JayF | so I'd also think for the general iscsi case | 01:04 |
| jroll | what do you mean "when using local config drive" | 01:04 |
| jroll | devananda: ^ | 01:04 |
| JayF | jroll: vs metadata config service ? | 01:04 |
| jroll | "local" being "not metadata"? | 01:04 |
| JayF | jroll: was my assumption at least | 01:04 |
| jroll | yeah, ok | 01:04 |
| devananda | ok. I now agree -- iscsi deploy driver should put config-2 partition at the tail of the drive | 01:04 |
| devananda | and still let autogrowfs handle expanding the root | 01:04 |
| JayF | FWIW the exact logic IPA uses --> https://github.com/openstack/ironic-python-agent/blob/master/ironic_python_agent/shell/copy_configdrive_to_disk.sh | 01:04 |
| devananda | jroll: I mean when not using virtualmedia-based-configdrive | 01:05 |
| devananda | or network-based | 01:05 |
| JayF | ooooh. | 01:05 |
| jroll | ok | 01:05 |
| devananda | I could mount one via NFS or iSCSI, for example | 01:05 |
| jroll | and now we're thinking iscsi should *always* create that partition? | 01:06 |
| devananda | (at least i think i could) | 01:06 |
| jroll | or just carve as needed | 01:06 |
| devananda | jroll: yes | 01:06 |
| devananda | ugh | 01:06 |
| devananda | either/or ... | 01:06 |
| jroll | heh | 01:06 |
| *** ChanServ sets mode: -o clarkb | 01:06 | |
| *** clarkb has left #openstack-ironic | 01:06 | |
| devananda | argh | 01:06 |
| devananda | I was about to say something and realized i'm wrong, again | 01:06 |
| devananda | it can't be a global config option, even though I think it should be | 01:07 |
| devananda | because E_UPGRADE | 01:07 |
| devananda | if I deploy a node today | 01:07 |
| devananda | then tomorrow I upgrade Ironic and turn on this feature | 01:07 |
| devananda | how do I rebuild? | 01:07 |
| devananda | so -- yes, I still think it should be a global option (because making this per-node is not necessary) | 01:07 |
| devananda | but deploy_utils needs to be able to detect nodes that were deployed without a config-2 partition and, when rebuilding, even if the current config option says "write a config-2 partition", it should not in those cases | 01:08 |
| * devananda stops thinking in text as it probably is confusing people | 01:08 | |
| jroll | hmm | 01:08 |
| NobodyCam | devananda: would global option prevent some setups where one zone (say prod) has a meta data service and another zone (cell) say dev is using a hand coded config for testing `something` | 01:10 |
| devananda | NobodyCam: a) that kind of config difference in prod vs. test is a terrible idea. b) different regions (which I think is what you mean here, not cell) would have different configs anyway | 01:11 |
| devananda | (I'm sure someone will try it, though) | 01:12 |
| NobodyCam | I never said it was a good thing | 01:12 |
| devananda | hehe | 01:12 |
| NobodyCam | :) | 01:12 |
| devananda | different regions have different config files | 01:13 |
| NobodyCam | just trying to think of a case where a global option owuld not work | 01:13 |
| NobodyCam | would* | 01:13 |
| devananda | by global, I mean a driver-specific config setting | 01:13 |
| devananda | rather than another option on node.driver_info | 01:14 |
| jroll | oh, wait, we want an ironic config as well? | 01:14 |
| devananda | I have not yet thought of a case where two nodes, using the same driver (eg, iscsi_ipmitool), would need to differ in this regard (one gets info from metadata service, the other from a configdrive partition) | 01:15 |
| openstackgerrit | Yuiko Takada proposed openstack/python-ironicclient: Fix node-set-provision-state cmd's help strings https://review.openstack.org/135518 | 01:15 |
| NobodyCam | oh driver-specific ... (in best gilda radner voice) never mind | 01:15 |
| jroll | I have not either | 01:15 |
| jroll | nova seems to support it | 01:15 |
| devananda | NobodyCam: ahh - sorry that I wasn't more specific! yes, if I had meant actually-really-global, then I would share your concern | 01:16 |
| devananda | jroll: oh? how is it selected? | 01:16 |
| NobodyCam | :) phew | 01:17 |
| * NobodyCam looks at clock and open beer! | 01:17 | |
| NobodyCam | opens even | 01:17 |
| JayF | NobodyCam: beer-thirty++ | 01:17 |
| jroll | 00:59:43 jroll | https://github.com/openstack/nova/blob/master/nova/virt/configdrive.py#L179-193 | 01:17 |
| NobodyCam | :) | 01:17 |
| jroll | devananda: ^ | 01:17 |
| devananda | urgh | 01:19 |
| devananda | it's an image property?! | 01:19 |
| jroll | ¯\_(ツ)_/¯ | 01:20 |
| jroll | I mean, it makes sense to some extent | 01:21 |
| jroll | one might want to usually use the metadata service | 01:21 |
| jroll | but then perhaps they have one image where dhcp is broken or whatever and they need to use configdrive in that case | 01:22 |
| devananda | damn. yes. | 01:22 |
| devananda | well, we should probably do the asme | 01:23 |
| devananda | same | 01:23 |
| jroll | I mean | 01:23 |
| jroll | I don't think ironic should have configs for this | 01:23 |
| jroll | if the nova instance needs a configdrive, it should be put in instance_info | 01:23 |
| jroll | and if ironic sees that key, it does it | 01:23 |
| devananda | i'm convinced | 01:24 |
| jroll | \o/ | 01:24 |
| jroll | we should put together some sort of process where we talk about implementation details through gerrit | 01:25 |
| *** achanda has quit IRC | 01:25 | |
| devananda | ..... | 01:25 |
| jroll | (I'm fairly certain this was all in the spec :P ) | 01:25 |
| devananda | there's a spec? | 01:25 |
| devananda | :P | 01:25 |
| jroll | though maybe the nova spec | 01:25 |
| jroll | ha | 01:25 |
| jroll | at any rate, with my last patch, it should work in IPA assuming nova does the right thing | 01:26 |
| jroll | I kind of want to do the pxe/ilo stuff if I can | 01:26 |
| jroll | though, now it occurs to me that ilo uses iscsi_deploy | 01:26 |
| jroll | which pxe also uses | 01:27 |
| jroll | but we want those implementations to be different | 01:27 |
| jroll | ugh | 01:27 |
| devananda | jroll: i did not see any of this in the spec I approved, fwiw | 01:27 |
| devananda | s/did/do/ | 01:28 |
| jroll | well, we didn't talk about driver implementation in the spec | 01:28 |
| jroll | :/ | 01:28 |
| devananda | 162 * Implement functionality for various deploy drivers. | 01:28 |
| jroll | right | 01:28 |
| devananda | right | 01:28 |
| jroll | I was talking about the nova/config stuff | 01:28 |
| devananda | yah. the details of where the partition is put down by the iscsi driver is not in that spec | 01:28 |
| *** penick has quit IRC | 01:28 | |
| devananda | which is, i think, where we started | 01:29 |
| jroll | right | 01:29 |
| jroll | yep | 01:29 |
| devananda | oh ya, i was warming up the button to update LP ... | 01:29 |
| *** chenglch has joined #openstack-ironic | 01:29 | |
| NobodyCam | jroll: you can always ref implementation details via irc logs (i.e. http://eavesdrop.openstack.org/irclogs/%23openstack-ironic/%23openstack-ironic.2014-11-20.log) | 01:33 |
| jroll | lol | 01:34 |
| jroll | Implements: blueprint http://eavesdrop.openstack.org/irclogs/%23openstack-ironic/%23openstack-ironic.2014-11-20.log | 01:34 |
| devananda | hah | 01:34 |
| NobodyCam | anything but another girrit page to go to | 01:34 |
| NobodyCam | gerrit :-p | 01:34 |
| jroll | ha | 01:35 |
| jroll | alright, I'm off for the night | 01:35 |
| jroll | see y'all tomorrow | 01:35 |
| NobodyCam | have a good night :) | 01:36 |
| devananda | hmm. gertty doesn't un-hide a review for which there is a new comment | 01:36 |
| *** dlaube has quit IRC | 01:36 | |
| jroll | before I go... | 01:43 |
| jroll | Haomeng|2: the spec is approved here now, please remove your -1 https://review.openstack.org/#/c/128388/ | 01:43 |
| Haomeng|2 | jroll: ok | 01:45 |
| Haomeng|2 | jroll: done | 01:46 |
| jroll | thanks :D | 01:46 |
| openstackgerrit | Merged openstack/ironic-specs: Allow short-specs https://review.openstack.org/133345 | 01:48 |
| *** imtiaz has quit IRC | 01:50 | |
| Haomeng|2 | jroll: yw | 01:56 |
| *** yjiang5 is now known as yjiang5_away | 01:57 | |
| *** alexpilotti has quit IRC | 01:58 | |
| *** nosnos has joined #openstack-ironic | 01:59 | |
| *** alexpilotti has joined #openstack-ironic | 02:00 | |
| *** alexpilotti has quit IRC | 02:01 | |
| *** alexpilotti has joined #openstack-ironic | 02:02 | |
| *** alexpilotti has quit IRC | 02:03 | |
| *** alexpilotti has joined #openstack-ironic | 02:04 | |
| *** alexpilotti has quit IRC | 02:06 | |
| *** rloo has quit IRC | 02:10 | |
| *** yuanying has quit IRC | 02:25 | |
| PaulCzar | curious why devstack's ironic apt list includes docker ? | 02:27 |
| PaulCzar | https://github.com/openstack-dev/devstack/blob/9b4b2913f9763bd7cfea11bfa68c84f95b5ff02f/files/apts/ironic#L1 | 02:27 |
| NobodyCam | PaulCzar: its a IPA thing | 02:31 |
| NobodyCam | lol | 02:31 |
| *** yuanying has joined #openstack-ironic | 02:35 | |
| Haomeng|2 | PaulCzar: yes, our IPA will run in docker | 02:36 |
| *** tonyb has joined #openstack-ironic | 02:36 | |
| openstackgerrit | jiangfei proposed openstack/ironic-python-agent: add deprecated_name to the config variables https://review.openstack.org/131632 | 02:47 |
| *** krtaylor has quit IRC | 03:12 | |
| *** HenryG has quit IRC | 03:15 | |
| *** HenryG has joined #openstack-ironic | 03:18 | |
| *** nosnos has quit IRC | 03:24 | |
| *** krtaylor has joined #openstack-ironic | 03:25 | |
| *** rushiagr_away is now known as rushiagr | 03:27 | |
| *** ramineni has joined #openstack-ironic | 03:29 | |
| *** naohirot has quit IRC | 03:33 | |
| *** nosnos has joined #openstack-ironic | 03:49 | |
| *** rushiagr is now known as rushiagr_away | 03:53 | |
| *** jgrimm is now known as zz_jgrimm | 03:54 | |
| *** yuanying_ has joined #openstack-ironic | 03:59 | |
| *** naohirot has joined #openstack-ironic | 04:00 | |
| *** yuanying has quit IRC | 04:03 | |
| *** spandhe has quit IRC | 04:08 | |
| PaulCzar | does that mean that devstack for ironic only supports fed20 / ubuntu 14 as earlier don't have native packages for docker ? | 04:09 |
| *** yuanying_ has quit IRC | 04:10 | |
| *** pcrews has quit IRC | 04:16 | |
| *** yuanying has joined #openstack-ironic | 04:16 | |
| *** yuanying has quit IRC | 04:23 | |
| *** pensu has joined #openstack-ironic | 04:23 | |
| *** rushiagr_away is now known as rushiagr | 04:33 | |
| *** yuanying has joined #openstack-ironic | 04:38 | |
| openstackgerrit | Merged openstack/python-ironicclient: Add 'API' to description of ironic command https://review.openstack.org/135794 | 04:42 |
| *** yuanying has quit IRC | 04:48 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic-specs: UEFI Secure Boot support for iLO drivers ============================================ https://review.openstack.org/135845 | 04:56 |
| *** yuanying has joined #openstack-ironic | 04:59 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic-specs: Ironic Management Interfaces to support UEFI Secure Boot ======================================================== https://review.openstack.org/135845 | 05:02 |
| *** pensu has quit IRC | 05:02 | |
| *** yuanying has quit IRC | 05:11 | |
| *** yuanying has joined #openstack-ironic | 05:14 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic-specs: UEFI Secure Boot support for iLO drivers https://review.openstack.org/135228 | 05:20 |
| *** yuanying has quit IRC | 05:26 | |
| *** yuanying has joined #openstack-ironic | 05:26 | |
| *** ryanpetrello has joined #openstack-ironic | 05:31 | |
| *** Hefeweizen has quit IRC | 05:33 | |
| *** Hefeweizen has joined #openstack-ironic | 05:34 | |
| *** achanda has joined #openstack-ironic | 05:38 | |
| *** yuanying has quit IRC | 05:41 | |
| *** pradipta_away is now known as pradipta | 05:44 | |
| *** penick has joined #openstack-ironic | 05:48 | |
| *** yuanying has joined #openstack-ironic | 05:48 | |
| *** penick_ has joined #openstack-ironic | 05:56 | |
| *** penick has quit IRC | 05:57 | |
| *** penick_ is now known as penick | 05:57 | |
| *** chenglch has quit IRC | 06:09 | |
| *** chenglch has joined #openstack-ironic | 06:23 | |
| openstackgerrit | Harshada Mangesh Kakad proposed openstack/ironic: Add serial console feature to seamicro driver https://review.openstack.org/132628 | 06:25 |
| *** harlowja is now known as harlowja_away | 06:31 | |
| *** imtiaz has joined #openstack-ironic | 06:33 | |
| *** k4n0 has joined #openstack-ironic | 06:33 | |
| *** ujuc has joined #openstack-ironic | 06:37 | |
| openstackgerrit | Tan Lin proposed openstack/ironic: Add driver_validate() https://review.openstack.org/134184 | 06:40 |
| *** ryanpetrello has quit IRC | 06:40 | |
| *** achanda has quit IRC | 06:45 | |
| *** pcrews has joined #openstack-ironic | 06:55 | |
| *** yuanying has quit IRC | 06:57 | |
| *** achanda has joined #openstack-ironic | 06:58 | |
| *** yuanying has joined #openstack-ironic | 06:58 | |
| *** achanda has quit IRC | 06:58 | |
| *** achanda has joined #openstack-ironic | 06:59 | |
| *** pcrews has quit IRC | 07:16 | |
| *** achanda has quit IRC | 07:25 | |
| *** achanda has joined #openstack-ironic | 07:25 | |
| *** achanda has quit IRC | 07:25 | |
| *** mrda is now known as mrda_away | 07:32 | |
| *** chenglch|2 has joined #openstack-ironic | 07:33 | |
| *** jcoufal has joined #openstack-ironic | 07:35 | |
| *** chenglch has quit IRC | 07:36 | |
| *** pensu has joined #openstack-ironic | 07:41 | |
| *** imtiaz has quit IRC | 07:46 | |
| *** imtiaz has joined #openstack-ironic | 07:47 | |
| *** imtiaz has quit IRC | 07:51 | |
| *** kbyrne has quit IRC | 07:55 | |
| *** kbyrne has joined #openstack-ironic | 08:00 | |
| *** chenglch has joined #openstack-ironic | 08:04 | |
| *** ifarkas has joined #openstack-ironic | 08:05 | |
| *** chenglch|2 has quit IRC | 08:07 | |
| *** dlpartain has joined #openstack-ironic | 08:09 | |
| *** jcoufal has quit IRC | 08:12 | |
| *** jcoufal has joined #openstack-ironic | 08:12 | |
| *** dtantsur|afk is now known as dtantsur | 08:14 | |
| dtantsur | Morning Ironic | 08:14 |
| *** Haomeng has joined #openstack-ironic | 08:26 | |
| *** Haomeng|2 has quit IRC | 08:26 | |
| *** pradipta is now known as pradipta_away | 08:31 | |
| *** pensu1 has joined #openstack-ironic | 08:39 | |
| *** pensu has quit IRC | 08:40 | |
| *** imtiaz has joined #openstack-ironic | 08:48 | |
| pensu1 | Morning dtantsur | 08:51 |
| *** imtiaz has quit IRC | 08:52 | |
| *** Haomeng|2 has joined #openstack-ironic | 08:54 | |
| *** Haomeng has quit IRC | 08:55 | |
| GheRivero | motning all | 08:56 |
| *** Nisha has joined #openstack-ironic | 09:00 | |
| dtantsur | pensu1, GheRivero, morning | 09:07 |
| *** derekh has joined #openstack-ironic | 09:11 | |
| *** ujuc has quit IRC | 09:11 | |
| *** Marga_ has quit IRC | 09:18 | |
| *** romcheg has joined #openstack-ironic | 09:27 | |
| *** lucasagomes has joined #openstack-ironic | 09:29 | |
| *** foexle has joined #openstack-ironic | 09:32 | |
| *** andreykurilin_ has joined #openstack-ironic | 09:34 | |
| romcheg | Good morning/evening guys! | 09:38 |
| romcheg | Have you seen this thread http://lists.openstack.org/pipermail/openstack-dev/2014-November/050979.html ? | 09:39 |
| romcheg | I think it's quite interesting | 09:39 |
| *** takadayu_ has joined #openstack-ironic | 09:40 | |
| ramineni | lucasagomes: hi | 09:41 |
| takadayu_ | Hi Ironic | 09:42 |
| ramineni | lucasagomes: need to discuss regarding your comment on https://review.openstack.org/#/c/132746/5/ironic/drivers/modules/ilo/management.py | 09:42 |
| lucasagomes | ramineni, hey afternoon :) | 09:44 |
| lucasagomes | ramineni, sure. Just took a look I think that instantianting it would be fine :) | 09:44 |
| lucasagomes | I understand it's a temp thing | 09:44 |
| lucasagomes | I will comment on the patch | 09:44 |
| lucasagomes | takadayu_, hello there! good afternoon! | 09:44 |
| lucasagomes | romcheg, yo morning | 09:44 |
| ramineni | lucasagomes : morning :) gr8 .. thanks | 09:45 |
| takadayu_ | lucasagomes, yeah, good evening! | 09:45 |
| lucasagomes | takadayu_, oh my bad (/me is bad with TZs) | 09:46 |
| *** MattMan has quit IRC | 09:46 | |
| lucasagomes | good evening :) | 09:46 |
| *** MattMan has joined #openstack-ironic | 09:47 | |
| *** chenglch has quit IRC | 09:47 | |
| takadayu_ | lucasagomes, never mind :D BTW, I'd like to commit a new patch fixing around https://github.com/openstack-dev/devstack/blob/master/lib/ironic#L587, is it OK? | 09:48 |
| *** imtiaz has joined #openstack-ironic | 09:49 | |
| *** imtiaz has quit IRC | 09:53 | |
| lucasagomes | takadayu_, oh yeah definitely :) thanks for that | 09:54 |
| takadayu_ | lucasagomes, thx, then I'll add you as a reviewer, so please review it if you have available time ;) | 09:54 |
| *** yuriyz has quit IRC | 09:55 | |
| lucasagomes | will do! | 09:55 |
| *** yuriyz has joined #openstack-ironic | 09:55 | |
| takadayu_ | I appreciate it :D | 09:59 |
| openstackgerrit | Anusha Ramineni proposed openstack/ironic: iLO Management Interface https://review.openstack.org/132746 | 10:01 |
| Nisha | dtantsur: lucasagomes hi | 10:02 |
| dtantsur | morning/afternoon/evening, romcheg, takadayu_, lucasagomes, Nisha, ramineni! | 10:03 |
| dtantsur | (so many of us :) | 10:03 |
| lucasagomes | :) indeed | 10:03 |
| lucasagomes | Nisha, hi | 10:03 |
| Nisha | dtantsur: i was going through your comments on ethe spec | 10:03 |
| Nisha | lucasagomes: i resolved the comments you gave yest, and update dthe spec | 10:04 |
| Nisha | please have a look | 10:04 |
| lucasagomes | Nisha, on the discover? sure | 10:04 |
| Nisha | dtantsur: i am actually confused on target_provision_state and provision_state now | 10:04 |
| Nisha | lucasagomes: yes | 10:04 |
| takadayu_ | dtantsur, good evening! I'll follow your greeting way :) | 10:04 |
| dtantsur | :) | 10:05 |
| *** arif-ali has quit IRC | 10:05 | |
| dtantsur | Nisha, "target" is a future state, that we aim to arrive at | 10:05 |
| *** imtiaz has joined #openstack-ironic | 10:05 | |
| Nisha | dtantsur: Ok , so in the start of the discovery, the provision state shall be set to DISCOVERING | 10:06 |
| Nisha | what shall be state of target_provision_state? | 10:06 |
| Nisha | since it is asynch in nature, i think it shall be wait_callback as it is today for deploy | 10:07 |
| dtantsur | Nisha, I think it's INIT | 10:07 |
| Nisha | INIT is final state , corect? | 10:07 |
| dtantsur | Nisha, yes, that's what target state means | 10:07 |
| Nisha | that is when discovery is completed | 10:07 |
| Nisha | and api has returned | 10:07 |
| *** andreykurilin_ has quit IRC | 10:08 | |
| dtantsur | wait_callback 1. will be deleted in favor of wait flag; 2. can't be target state. Users do not request node to go into wait_callback :) | 10:08 |
| dtantsur | Nisha, right. so it's our target during discovery | 10:08 |
| dtantsur | lucasagomes, ^^^ am I correct? | 10:08 |
| * lucasagomes reads | 10:08 | |
| *** penick has quit IRC | 10:09 | |
| *** andreykurilin_ has joined #openstack-ironic | 10:09 | |
| Nisha | so target_provision_state shall be None when discovery is just initiated? | 10:09 |
| *** imtiaz has quit IRC | 10:09 | |
| Nisha | or just before invoking discovery when provision_state is set to DISCOVERING | 10:10 |
| lucasagomes | dtantsur, yup sounds correc | 10:10 |
| lucasagomes | correct* | 10:10 |
| dtantsur | Nisha, again, target is some valid state. None won't be a valid state, neither will be wait_callback | 10:11 |
| Nisha | dtantsur: then? | 10:11 |
| dtantsur | Nisha, where are you going to get after discovery? To INIT. Then target state should be INIT. | 10:11 |
| Nisha | dtantsur: 1. just before invoking discovery: | 10:12 |
| lucasagomes | Nisha, the state machine still being discussed. Right now the way it goes is. ENROLL->DISCOVERING->INIT | 10:12 |
| lucasagomes | so machines starts at ENROLL, if discover is enabled by setting target_state to INIT | 10:12 |
| lucasagomes | means it goes first to DISCOVERING (which is a transient state) to INIT | 10:13 |
| Nisha | provision_state is set to DISCOVERING, and target_provison_state = ? | 10:13 |
| lucasagomes | I personally don't like that flow (see my suggestions on the state machine spec) | 10:13 |
| lucasagomes | Nisha, target is set to the next non-transitional ('ing) state | 10:13 |
| dtantsur | for any transient state, target state is a valid state we want to get into | 10:13 |
| dtantsur | for any non-transient state, target state don't exists and is set to None | 10:14 |
| Nisha | could you explain 'non-transient' here | 10:14 |
| dtantsur | that said, if we want to get into INIT from discovery, target state during discovery is INIT | 10:14 |
| Nisha | ok i got | 10:14 |
| Nisha | understood | 10:14 |
| dtantsur | Nisha, non-transient is a state in which node stays as long as operator wants | 10:14 |
| Nisha | let me just repeat what i understood | 10:15 |
| dtantsur | Nisha, transient state is temporary and will be changed by Ironic (or 3rdparty service), once some operation is ended | 10:15 |
| Nisha | 1. when discovery started : provision_state = DISCOVERING and target_provision_state=INIT | 10:15 |
| Nisha | 2. discovery done: provision_state = INIT and target_provision_state = None | 10:16 |
| Nisha | above is in case of success | 10:16 |
| lucasagomes | Nisha, (off-topic) would also be good to get ur input on the state machine spec, we might be overcomplicating some of it (I think we are, that's why I don't like the ENROLL -> INIT state) | 10:16 |
| Nisha | i actually went thru the spec | 10:16 |
| *** arif-ali has joined #openstack-ironic | 10:16 | |
| Nisha | lucasagomes: i will do | 10:17 |
| lucasagomes | Nisha, thanks :) | 10:17 |
| Nisha | lucasagomes: thanks :) | 10:17 |
| Nisha | lucasagomes: dtantsur could you confirm if my understanding is correct? | 10:17 |
| Nisha | on states thing | 10:17 |
| dtantsur | Nisha, looks correct | 10:17 |
| lucasagomes | Nisha, sounds correct | 10:17 |
| Nisha | thanks. will update the spec, | 10:18 |
| Nisha | any other comments on the spec are welcome | 10:18 |
| Nisha | lucasagomes: dtantsur request your reviews on https://review.openstack.org/134022 also | 10:20 |
| Nisha | raised it a week back | 10:20 |
| openstackgerrit | Imre Farkas proposed openstack/ironic-specs: New driver interface for RAID configuration https://review.openstack.org/135899 | 10:21 |
| dtantsur | yeah, it's on the radar | 10:21 |
| Nisha | :) | 10:21 |
| dtantsur | I have >10 patches to review this morning already :( | 10:21 |
| Nisha | :) ok. Core reviewiers job is that :) | 10:22 |
| Nisha | dtantsur: lucasagomes one more ques i had | 10:23 |
| Nisha | for some packages which are required to be installed on conductor but are not available on pypi, but thru apt-get how does ironic go about with that | 10:24 |
| *** naohirot has quit IRC | 10:25 | |
| dtantsur | Nisha, I guess it's distro maintainers' job to make these dependencies/recommendations | 10:27 |
| lucasagomes | Nisha, third-party? If not mandatory for Ironic I'm fine with that, just need to document that operator should install it on the image (just like they should install pyghmi, wsman, etc... if they are willing to use ipminative or drac) | 10:27 |
| Nisha | dtantsur: lucasagomes i am talking about packages like grub, mkisofs required by ironic drivers | 10:28 |
| Nisha | for grub, the latest beta version is buggy | 10:28 |
| Nisha | the stable version is supported only on precise ubuntu | 10:29 |
| dtantsur | I guess it's again distro's job to make a working version available | 10:29 |
| Nisha | :) | 10:29 |
| dtantsur | Ironic can't force Ubuntu not to break things :) | 10:29 |
| Nisha | dtantsur: :0 | 10:29 |
| Nisha | :) | 10:29 |
| Nisha | ok. | 10:29 |
| lucasagomes | Nisha, it's something like ipmitool is for the ipmitool driver right? | 10:29 |
| Nisha | yes | 10:30 |
| lucasagomes | right, so I think it would be the same for these other tools etc | 10:30 |
| lucasagomes | ipmitool does some version checks to see what is supported and what is not | 10:30 |
| lucasagomes | and adapt to it etc | 10:30 |
| dtantsur | it's usually like: we document the dependencies, distro maintainers package Ironic and dependencies and make sure they work together | 10:30 |
| lucasagomes | or fail | 10:30 |
| *** dtantsur is now known as dtantsur|bbl | 10:33 | |
| *** pelix has joined #openstack-ironic | 10:34 | |
| *** pensu has joined #openstack-ironic | 10:39 | |
| Nisha | lucasagomes: ok | 10:40 |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic-specs: Discover node properties using new CLI node-discover-properties https://review.openstack.org/100951 | 10:40 |
| Nisha | lucasagomes: dtantsur|bbl ^^^^ please review | 10:41 |
| Nisha | lucasagomes: resolved the comments | 10:41 |
| *** pensu1 has quit IRC | 10:41 | |
| lucasagomes | cool thanks I will take a look soonish | 10:42 |
| *** Nisha has quit IRC | 10:45 | |
| openstackgerrit | Tan Lin proposed openstack/ironic: Add driver_validate() https://review.openstack.org/134184 | 10:47 |
| *** igordcard has joined #openstack-ironic | 10:48 | |
| *** takadayu_ has quit IRC | 10:50 | |
| *** sambetts has joined #openstack-ironic | 10:54 | |
| *** ramineni has quit IRC | 11:05 | |
| *** imtiaz has joined #openstack-ironic | 11:06 | |
| *** zz_jgrimm is now known as jgrimm | 11:06 | |
| *** Haomeng has joined #openstack-ironic | 11:10 | |
| *** imtiaz has quit IRC | 11:10 | |
| *** Haomeng|2 has quit IRC | 11:11 | |
| *** jcoufal has quit IRC | 11:11 | |
| *** jcoufal has joined #openstack-ironic | 11:12 | |
| *** Haomeng|2 has joined #openstack-ironic | 11:15 | |
| *** Haomeng has quit IRC | 11:15 | |
| *** naohirot has joined #openstack-ironic | 11:21 | |
| *** jcoufal has quit IRC | 11:27 | |
| *** jcoufal has joined #openstack-ironic | 11:27 | |
| *** andreykurilin_ has quit IRC | 11:27 | |
| *** andreykurilin_ has joined #openstack-ironic | 11:28 | |
| *** imtiaz has joined #openstack-ironic | 12:06 | |
| *** kbyrne has quit IRC | 12:09 | |
| *** imtiaz has quit IRC | 12:11 | |
| *** rushiagr is now known as rushiagr_away | 12:15 | |
| *** andreykurilin_ has quit IRC | 12:15 | |
| *** aszeszo has joined #openstack-ironic | 12:16 | |
| *** dlpartain has quit IRC | 12:23 | |
| *** dlpartain has joined #openstack-ironic | 12:33 | |
| *** jcoufal has quit IRC | 12:42 | |
| *** jcoufal has joined #openstack-ironic | 12:43 | |
| *** rushiagr_away is now known as rushiagr | 12:46 | |
| *** jcoufal has quit IRC | 12:51 | |
| *** Shrews_ has joined #openstack-ironic | 12:58 | |
| *** pensu has quit IRC | 12:59 | |
| *** dtantsur|bbl is now known as dtantsur | 12:59 | |
| *** jcoufal has joined #openstack-ironic | 13:01 | |
| *** nosnos has quit IRC | 13:05 | |
| *** nosnos has joined #openstack-ironic | 13:06 | |
| *** imtiaz has joined #openstack-ironic | 13:07 | |
| *** imtiaz has quit IRC | 13:11 | |
| toabctl | can anybody point me to the bugtracker of pyghmi ? | 13:12 |
| *** lazy_prince is now known as killer_prince | 13:14 | |
| dtantsur | toabctl, I haven't found one last time :( so you may want to contact it's devs directly | 13:15 |
| toabctl | dtantsur: the question is who I should contact... | 13:15 |
| dtantsur | toabctl, judging by https://github.com/stackforge/pyghmi/commits/master it's Jarrod Johnon, but I don't know the contacts.. | 13:17 |
| dtantsur | toabctl, try this: http://sourceforge.net/p/xcat/wiki/XCAT_Support/ | 13:18 |
| toabctl | dtantsur: seems that mordred also contributed to the project. | 13:18 |
| toabctl | mordred: do you know if there's a bugtracker for pyghmi? and if so, where? | 13:19 |
| dtantsur | toabctl, actually it's a good question to raise on a ML. you're at least the second to ask this question recently | 13:19 |
| toabctl | dtantsur: there are also no git tags for the latest releases on github | 13:20 |
| dtantsur | oh, that's messy... | 13:22 |
| dtantsur | lucasagomes, any ideas ^^^ | 13:22 |
| Haomeng|2 | toabctl: pyghmi is from xcat, so the bug list you want to see that should be - http://sourceforge.net/p/xcat/bugs/ | 13:22 |
| toabctl | Haomeng|2: searching for pyghmi in the buglist brings 0 bugs. there is also no component for pyghmi . I'm not sure if that's the bugtracker. | 13:26 |
| toabctl | Haomeng|2: and the code seems to be on github, not on stackforge | 13:26 |
| Haomeng|2 | toabctl: ok, let me check with xcat team | 13:26 |
| Haomeng|2 | toabctl: yes | 13:26 |
| toabctl | Haomeng|2: ok. thx | 13:26 |
| Haomeng|2 | toabctl: should be owned by xcat | 13:27 |
| dtantsur | Haomeng|2, please make the xcat team aware that people regularly struggle with finding whom and how to contact about pyghmi | 13:28 |
| dtantsur | it should be better communicated IMO | 13:28 |
| Haomeng|2 | dtantsur: ok | 13:28 |
| Haomeng|2 | dtantsur: will check with xcat team members | 13:28 |
| *** athomas has joined #openstack-ironic | 13:28 | |
| Haomeng|2 | dtantsur: to find the owner | 13:28 |
| dtantsur | thanks! | 13:28 |
| Haomeng|2 | dtantsur: yw | 13:30 |
| *** andreykurilin_ has joined #openstack-ironic | 13:30 | |
| *** jiangfei has quit IRC | 13:32 | |
| *** jiangfei has joined #openstack-ironic | 13:33 | |
| *** dlpartain has left #openstack-ironic | 13:33 | |
| jroll | morning ironic :) | 13:34 |
| jroll | PaulCzar: right, docker is a build dependency for ironic with the agent driver... you can rip that dep out if you're using 12.04 and not building agent ramdisks :) | 13:34 |
| jroll | dtantsur: I thought pyghmi just shared a bug tracker with us, imbw | 13:34 |
| *** k4n0 has quit IRC | 13:38 | |
| *** andreykurilin__ has joined #openstack-ironic | 13:40 | |
| *** andreykurilin_ has quit IRC | 13:40 | |
| NobodyCam | morning Ironic, says the man slowly drinking coffee | 13:44 |
| *** dprince has joined #openstack-ironic | 13:47 | |
| sambetts | o/ | 13:47 |
| NobodyCam | morning sambetts :) | 13:47 |
| dtantsur | jroll, morning :) I don't think so, our teams look different | 13:52 |
| jroll | heya NobodyCam sambetts | 13:52 |
| dtantsur | NobodyCam, sambetts, morning! | 13:52 |
| jroll | and dtantsur :) | 13:52 |
| jroll | yeah, imbw, but I swear I've seen pyghmi bugs there | 13:53 |
| NobodyCam | huh /me thinks something is up! https://pypi.python.org/pypi/ironic <- is that blank for others too? | 13:53 |
| NobodyCam | morning dtantsur jroll | 13:53 |
| *** andreykurilin__ has quit IRC | 13:53 | |
| dtantsur | NobodyCam, O_O | 13:54 |
| dtantsur | yes | 13:54 |
| NobodyCam | humm | 13:54 |
| jroll | huh | 13:54 |
| jroll | do we publish to pypi? (I assume we do) | 13:54 |
| Haomeng|2 | NobodyCam: I checked, others works fine | 13:54 |
| jroll | nova hasn't published since 2013.1 | 13:55 |
| dtantsur | jroll, special -infra job does | 13:55 |
| jroll | https://pypi.python.org/pypi/glance 404s | 13:55 |
| dtantsur | so it's a question for #infra IMO | 13:55 |
| NobodyCam | ahh maybe we only publish the client | 13:55 |
| jroll | how sure are you about that? | 13:55 |
| jroll | I don't see other projects publishing | 13:55 |
| NobodyCam | https://pypi.python.org/pypi/python-ironicclient | 13:55 |
| jroll | https://pypi.python.org/pypi/swift 404 | 13:55 |
| jroll | yeah, don't think it's a problem | 13:56 |
| NobodyCam | so ya just my lack of coffee | 13:56 |
| NobodyCam | sorry for the scare :-p | 13:56 |
| dtantsur | ah yeah | 13:56 |
| NobodyCam | morning Haomeng|2 :) | 13:56 |
| jroll | ha | 13:56 |
| Haomeng|2 | NobodyCam: morning:) | 13:56 |
| dtantsur | and we have some different Keystone there https://pypi.python.org/pypi/Keystone :D | 13:56 |
| Haomeng|2 | NobodyCam: I am in evening:) | 13:57 |
| NobodyCam | hehehe its always morning (somewhere) /me is just waiting for the sun to come up | 13:57 |
| Haomeng|2 | yes, others works fine, but our ironicclient in pypi is blank | 13:57 |
| NobodyCam | heheh | 13:57 |
| NobodyCam | client is blank? | 13:58 |
| NobodyCam | or just Ironic it self | 13:58 |
| NobodyCam | https://pypi.python.org/pypi/python-ironicclient looks ok to me | 13:58 |
| *** rushiagr is now known as rushiagr_away | 13:59 | |
| *** rloo has joined #openstack-ironic | 14:01 | |
| NobodyCam | Hey dtantsur while your here, your agenst the readme update that adds the help output to the readme? are you thinking there is another maybe better way to provide that info? | 14:02 |
| dtantsur | NobodyCam, yes, an informal description of the most popular commands and link to `ironic help` or online docs for the details | 14:02 |
| *** jistr has joined #openstack-ironic | 14:02 | |
| NobodyCam | ok I will see what I can comeup with today :) | 14:03 |
| *** jistr is now known as jistr|mtg | 14:04 | |
| rloo | hello ironickers | 14:05 |
| NobodyCam | good mornig rloo :) | 14:05 |
| rloo | NobodyCam: wrt the readme for the client, I don't know why, but I think it is useful to know what commands are available per client release. | 14:05 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic: Updated from global requirements https://review.openstack.org/135963 | 14:05 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/ironic-python-agent: Updated from global requirements https://review.openstack.org/135964 | 14:05 |
| NobodyCam | rloo: I do too | 14:06 |
| jroll | morning rloo | 14:06 |
| dtantsur | rloo, hi | 14:06 |
| rloo | NobodyCam: but whether it is actually in the readme, or we put it on a wiki and link from the readme, I don't know. | 14:06 |
| jroll | rloo: NobodyCam: you know we're going to forget to update that :| | 14:06 |
| rloo | afternoon dtantsur, morning jroll. | 14:06 |
| rloo | jroll: yeah, that's the problem. | 14:07 |
| *** rushiagr_away is now known as rushiagr | 14:07 | |
| jroll | I don't see a problem with folks just running "ironic help" | 14:07 |
| NobodyCam | how often do we cut pypi releases | 14:07 |
| rloo | jroll: although, i have on my list, to add a link to the readme that points to the changes. (I don't recall the link to the changes right now...) | 14:07 |
| *** imtiaz has joined #openstack-ironic | 14:08 | |
| jroll | yes, I like that | 14:08 |
| rloo | NobodyCam: so far, we've cut a release at least twice? per openstack cycle. I don't know but you could look I guess. sec. | 14:09 |
| jroll | this has two +2's if anyone wants to look :) https://review.openstack.org/#/c/135792/ | 14:10 |
| jroll | rloo: we cut clients more often than release cycles | 14:10 |
| NobodyCam | the info I was looking for was what commands can "this (pypi) release of the cli do" and a link to a change may provide changes that are not included in the current pypi release as was already pointed out on the current review | 14:10 |
| rloo | NobodyCam: I'm not sure how to find out the various versions (it isn't obvious from https://pypi.python.org/pypi/python-ironicclient | 14:11 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/python-ironicclient: Updated from global requirements https://review.openstack.org/135985 | 14:11 |
| rloo | so far, I think devananda has been doing the releases, so maybe we should discuss with him, if something ought to be updated at release time. | 14:12 |
| dtantsur | I'm not convinced that README should be a source of "what this release can do". I always though of a README as a starting point for a user. | 14:12 |
| *** imtiaz has quit IRC | 14:12 | |
| jroll | yeah, I believe we have a changelog, we should use it | 14:12 |
| rloo | dtantsur: I know that when I've looked at other packages on pypi.python.org, I've been annoyed at the lack of docn available for the packages | 14:13 |
| dtantsur | rloo, lack of docs - yeah. lack of a simple README too. | 14:13 |
| dtantsur | well, for ironic-discoverd a have a changelog in the end of README | 14:13 |
| rloo | dtantsur: so do you think the current readme is 'simple'? | 14:13 |
| dtantsur | rloo, current one is too simple :) lemme find some good example... | 14:14 |
| dtantsur | rloo, this is nice: https://pypi.python.org/pypi/testtools/ | 14:15 |
| dtantsur | it has project goals, examples. It only lacks a link to real docs :) | 14:15 |
| rloo | dtantsur: i'm not sure the client needs a 'why use it' section? | 14:16 |
| rloo | dtantsur: do you think we should give an example? if so, of what? | 14:16 |
| dtantsur | rloo, of the basic operations with client. create node, create port, list nodes, list ports for nodes, show a node... | 14:17 |
| dtantsur | to be kind of quick start guide | 14:17 |
| rloo | dtantsur: do you think it should be in the readme, or is it OK to have a link to that info? | 14:17 |
| NobodyCam | jroll: where is our client change-log? | 14:17 |
| rloo | NobodyCam: I just found it: http://git.openstack.org/cgit/openstack/python-ironicclient/tag/?id=0.1.3 | 14:18 |
| NobodyCam | dtantsur: actual code examples? | 14:18 |
| dtantsur | rloo, link should lead to comprehensive docs. and this is just quick start to get some idea | 14:18 |
| openstackgerrit | Jim Rollenhagen proposed openstack/python-ironicclient: Add maintenance_reason to node-show output https://review.openstack.org/135999 | 14:18 |
| jroll | NobodyCam: I just assumed we had one :) | 14:18 |
| jroll | speaking of the client, easy change right there that I overlooked | 14:18 |
| dtantsur | NobodyCam, a couple of CLI examples, a couple of code | 14:18 |
| rloo | jroll: ^^ the change-log. here it is again: http://git.openstack.org/cgit/openstack/python-ironicclient/tag/?id=0.1.3 | 14:19 |
| jroll | rloo: yep, saw that | 14:19 |
| NobodyCam | http://git.openstack.org/cgit/openstack/python-ironicclient/refs | 14:20 |
| jroll | right | 14:20 |
| NobodyCam | ok I'll refactor that patch today | 14:20 |
| rloo | NobodyCam: cool. | 14:20 |
| NobodyCam | til | 14:20 |
| rloo | dtantsur: just examples of CLI, or also examples of using the library? | 14:21 |
| dtantsur | at least CLI | 14:21 |
| jroll | we just ned quick examples | 14:21 |
| jroll | we have docs for the rest | 14:21 |
| NobodyCam | thank y'all for the input :) | 14:21 |
| rloo | we currently *only* have quick examples :-( | 14:22 |
| jroll | oh | 14:23 |
| jroll | well, we should fix our docs | 14:23 |
| rloo | NobodyCam: what about having a wiki with the ironic commands per release. something similar to https://wiki.openstack.org/wiki/Ironic/ReleaseNotes/Juno | 14:23 |
| jroll | we have a changelog, why not just use that? | 14:24 |
| rloo | jroll: yeah, we should. Some sort of usage/operator guide? | 14:24 |
| jroll | well | 14:24 |
| jroll | we have "ironic help" for that | 14:24 |
| jroll | we sohuld have better python library docs, though | 14:25 |
| jroll | speaking of which | 14:25 |
| jroll | was talking to another team at rackspace who is playing with ironic | 14:25 |
| jroll | and they said our docs are horrible :( | 14:25 |
| rloo | did they mention what was horrible about it, what needs to be improved? | 14:25 |
| rloo | just 'horrible' doesn't help much | 14:25 |
| jroll | all the things, really | 14:26 |
| jroll | they'll be lurking here soon asking questions and whatnot | 14:27 |
| rloo | maybe they could help by filing bugs then | 14:27 |
| NobodyCam | I agree, I've been trying to poke folks for info on what we can do to make it better | 14:27 |
| NobodyCam | s/it/them/ | 14:27 |
| jroll | rloo: yeah, they will be | 14:28 |
| rloo | i don't agree that our doc is incomplete. i think 'horrible' seems extreme though. | 14:28 |
| jroll | yeah | 14:28 |
| sambetts | I agree, horible is a bit harsh | 14:28 |
| jroll | that might be my word | 14:28 |
| jroll | or talking about agent docs :( | 14:29 |
| jroll | this was a few days ago, I'm low on RAM | 14:29 |
| *** enterprisedc_ has joined #openstack-ironic | 14:29 | |
| NobodyCam | uless he was talking about the IPA docs | 14:29 |
| jroll | ^ | 14:29 |
| * NobodyCam *ducks* | 14:29 | |
| jroll | lol | 14:29 |
| NobodyCam | hehehee | 14:29 |
| rloo | speaking of which, i should fix that user-guide page. today. maybe. | 14:29 |
| *** enterprisedc has quit IRC | 14:30 | |
| *** enterprisedc_ is now known as enterprisedc | 14:30 | |
| rloo | can't really complain about IPA docs, there aren't any? ;) | 14:30 |
| NobodyCam | lol | 14:30 |
| jroll | :| | 14:30 |
| * jroll is a bad person | 14:30 | |
| Haomeng|2 | NobodyCam: :) | 14:30 |
| NobodyCam | rloo: there is https://wiki.openstack.org/wiki/Ironic-python-agent | 14:31 |
| NobodyCam | lol | 14:31 |
| NobodyCam | jroll: not! | 14:31 |
| jroll | so old | 14:31 |
| jroll | so so so old | 14:31 |
| NobodyCam | :-p | 14:31 |
| rloo | NobodyCam: oh right, I forgot about that! is it horrible though? ;) | 14:31 |
| NobodyCam | lol | 14:32 |
| rloo | jroll: you're making me feel old... | 14:32 |
| jroll | ha! | 14:32 |
| * NobodyCam is old... and feels like it ...lol | 14:32 | |
| jroll | so I'm not making any promises, but I would love to hack on putting this in the IPA ramdisk tomorrow: https://github.com/openstack/diskimage-builder/blob/master/elements/deploy-ironic/init.d/80-deploy-ironic | 14:33 |
| jroll | and switch between the IPA and iSCSI container with a kernel param | 14:33 |
| NobodyCam | jroll: oh | 14:33 |
| NobodyCam | ++++ | 14:34 |
| *** nosnos has quit IRC | 14:34 | |
| *** ryanpetrello has joined #openstack-ironic | 14:40 | |
| jroll | NobodyCam: do you know where all of those env variables get set? e.g. ISCSI_TARGET_IQN | 14:42 |
| * jroll checks with tripleo as well | 14:43 | |
| *** r-daneel has joined #openstack-ironic | 14:46 | |
| *** bradjones has joined #openstack-ironic | 14:46 | |
| NobodyCam | jroll: here?? https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/iscsi_deploy.py#L337-L343 | 14:52 |
| jroll | lol | 14:52 |
| jroll | I mean in the ramdisk | 14:52 |
| *** mjturek has joined #openstack-ironic | 14:52 | |
| jroll | but I found it... https://github.com/openstack/diskimage-builder/blob/master/elements/ramdisk-base/init.d/20-init-variables | 14:52 |
| NobodyCam | ahh | 14:52 |
| NobodyCam | :-p | 14:52 |
| * NobodyCam wants to be evil and include "ironic chassis-create -d 'test chassis'" as an example command ... because its used so often | 14:53 | |
| jroll | lol | 14:54 |
| jroll | I've never touched chassis | 14:54 |
| NobodyCam | i know.. I had a great use case and deva shot me down at the summit | 14:55 |
| NobodyCam | lol | 14:55 |
| jroll | lol | 14:56 |
| jroll | what use case? | 14:56 |
| NobodyCam | standalone ironic, two nodes with all deploy/driver info added to a chassis, then operator just issues 'ironic chassis-set-power-state on' and both nodes power on and deploy with a single command | 14:58 |
| *** kbyrne has joined #openstack-ironic | 14:59 | |
| jroll | NobodyCam, JayF: https://bugs.launchpad.net/ironic/+bug/1394609 small outline of how this dual ramdisk would work | 14:59 |
| jroll | NobodyCam: interesting, that's cool | 15:00 |
| NobodyCam | there are many holes in that | 15:00 |
| jroll | yeah, but I like the concept | 15:00 |
| NobodyCam | lol :-p | 15:00 |
| * lucasagomes scrollsback :D | 15:00 | |
| openstackgerrit | Roman Prykhodchenko proposed openstack/ironic: Refactoring for Ironic policy https://review.openstack.org/126265 | 15:01 |
| jroll | lucasagomes: tl;dr all-in-one iscsi/agent ramdisk | 15:01 |
| NobodyCam | morning lucasagomes | 15:01 |
| * NobodyCam makes a more coffee | 15:01 | |
| rloo | anyone know what Harshada's nick is? or if they're online now? | 15:01 |
| lucasagomes | NobodyCam, morning | 15:02 |
| lucasagomes | rloo, nop | 15:02 |
| lucasagomes | jroll, oh that would be good | 15:02 |
| rloo | afternoon lucasagomes | 15:02 |
| lucasagomes | jroll, cause I just saw the link for the boot_options to build the template | 15:02 |
| lucasagomes | on the iPXE driver I kinda gotta refactor that out from PXE, actually many of the functions needs to go to a common place | 15:02 |
| jroll | lucasagomes: oh, yeah, basically going to make a container that runs those bash scripts | 15:02 |
| lucasagomes | cause they are reusable | 15:02 |
| jroll | yeah | 15:03 |
| rloo | lucasagomes: i had two questions about the vendor stuff. I noticed that this page mentions the passthrus: http://docs.openstack.org/developer/ironic/dev/drivers.html | 15:03 |
| lucasagomes | rloo, right... what are the questions? | 15:04 |
| rloo | lucasagomes: maybe http://docs.openstack.org/developer/ironic/dev/vendor-passthru.html needs to be combined? | 15:04 |
| lucasagomes | rloo, ohhh | 15:04 |
| * lucasagomes reads | 15:04 | |
| rloo | lucasagomes: 2nd question. wrt the decorators @passthru, @driver_passthru. Do people think those are clear? | 15:05 |
| lucasagomes | rloo, indeed the differenciation of node and driver are duplicated in our docs | 15:05 |
| rloo | lucasagomes: is @node_passthru clearer than @passthru? it seems like they're both driver passthrus, except one requires a node... | 15:05 |
| lucasagomes | rloo, I remember you proposed @node_passthru and @driver_passthru | 15:05 |
| lucasagomes | I like it to be more explicit, so I'm +1 for the idea | 15:06 |
| jroll | but now there's backward compatibility issues :| | 15:07 |
| rloo | lucasagomes: if we change to node_passthru, driver_passthru, is that clear to folks? we're so used to it now. (I'm not sure what else to name them though.) driver_node_passthru? node_driver_passthru? | 15:07 |
| jroll | if we change it now we could theoretically break out of tree drivers | 15:07 |
| rloo | no backward compatibility for the decorators. I don't think we need to worry about backwards w/i a cycle. and there were no decorators in juno. | 15:08 |
| jroll | you don't think there's anyone continuously deploying ironic? | 15:08 |
| lucasagomes | rloo, I think people are constantly deploying from master | 15:08 |
| lucasagomes | maybe we need to do a passthru = node_passthru | 15:08 |
| lucasagomes | like an alias with the decorator reference | 15:08 |
| rloo | no, i don't disagree that there might be folks deploying from master, I just don't think we ought to worry about backwards compatibility w/i master w/i a cycle. | 15:08 |
| *** imtiaz has joined #openstack-ironic | 15:08 | |
| lucasagomes | just in case, it's a 1 line code | 15:08 |
| rloo | eg, if we did, we wouldn't be able to land patches in a series w/o landing them all at once. | 15:09 |
| jroll | no, we should worry about it within a cycle | 15:09 |
| rloo | but if that is a concern, let me know. I would have -1'd lucas' patch when I first mentioned it. | 15:09 |
| jroll | and it's easy to do, alias passthru to node_passthru with a deprecation warning | 15:09 |
| lucasagomes | rloo, right actually it was on the logging patch | 15:10 |
| openstackgerrit | Naohiro Tamura proposed openstack/ironic-specs: iRMC Management Driver for Ironic https://review.openstack.org/136020 | 15:10 |
| lucasagomes | maybe I should have named then node_passthru and driver_passthru | 15:10 |
| rloo | in this case, it is easy to do. my concern is having to worry about breaking backwards compatibility w/i a cycle. that is something i'd have to keep in mind for all patches and that seems like a lot to remember to review for. | 15:11 |
| lucasagomes | once I started differentiating | 15:11 |
| lucasagomes | rloo, we should try our best, but indeed it's hard | 15:11 |
| jroll | rloo: I think we should, but I'm not a dictator | 15:11 |
| jroll | we shouldn't worry about it *as* hard | 15:11 |
| jroll | as between cycles | 15:11 |
| rloo | so in this case, we should decide now (before it gets harder) if we want to change the passthru decorator names, and if so, to what. | 15:11 |
| jroll | I think passthru is fine, I also think node_passthru is fine | 15:12 |
| rloo | wrt backwards compatibility w/i a cycle -- that seems like a general question for all openstack projects then? | 15:12 |
| lucasagomes | they seems to do it, at least projects like Nova | 15:12 |
| jroll | just because not all projects are awesome doesn't mean we can't be awesome :) | 15:12 |
| lucasagomes | :P | 15:12 |
| jroll | I think all projects should do it, but that's their decision | 15:13 |
| lucasagomes | we even have a mascot now! | 15:13 |
| rloo | jroll: touche. | 15:13 |
| jroll | \o/ | 15:13 |
| * lucasagomes needs to send out the pool for the name | 15:13 | |
| *** imtiaz has quit IRC | 15:13 | |
| *** linggao has joined #openstack-ironic | 15:13 | |
| lucasagomes | btw I was tinking about filtering the names, until we get 5 and then send out the pool (but it's out of topic) | 15:13 |
| rloo | jroll: is that something to bring up in the ML then? how many folks here are for backwards compatibility w/i a cycle? | 15:13 |
| lucasagomes | rloo, jroll so @passthru or @node_passthru | 15:13 |
| lucasagomes | I don't mind either, I think @node_passthru might be slightly better | 15:14 |
| rloo | node_driver_passthru? | 15:14 |
| lucasagomes | too long | 15:14 |
| jroll | lucasagomes: I honestly don't care which, node_passthru is slightly better but is it worth it? | 15:14 |
| lucasagomes | (imo) | 15:14 |
| jroll | it's just code | 15:14 |
| jroll | there are docs on what they dp | 15:14 |
| jroll | do | 15:14 |
| rloo | i think it is worth thinking about it. it isn't 'just code', its sort of an api/interface with drivers. | 15:14 |
| lucasagomes | jroll, yeah idk if it's worth, it. I think it's better cause it's more explicity and in the docs we say "Node VendorPassthru vs Driver VendorPassthru" | 15:14 |
| jroll | yeah | 15:15 |
| lucasagomes | but either name are sane I believe | 15:15 |
| lucasagomes | (I would rename preserver_ephemeral for e.g) | 15:15 |
| *** jcoufal has quit IRC | 15:15 | |
| lucasagomes | that's ^^^^ bad | 15:15 |
| lucasagomes | preserve_* | 15:15 |
| jroll | 15:13:54 rloo | jroll: is that something to bring up in the ML then? how many folks here are for backwards compatibility w/i a cycle? <- yes | 15:16 |
| jroll | lucasagomes: ha yeah | 15:16 |
| *** pcrews has joined #openstack-ironic | 15:17 | |
| lucasagomes | rloo, are u putting something in the ML? | 15:18 |
| rloo | i just don't see how we can do development if we preserve backwards compatibility. eg, the passthrus changed wrt their arguments, and ended up with the http-method having to be specified. that would have broken any drivers that used the older version of the passthru. | 15:18 |
| lucasagomes | I think it worths too | 15:18 |
| lucasagomes | some people might say we should do on releases, maybe milestones | 15:19 |
| rloo | lucasagomes: I could I guess... Ok, I'll write something up. | 15:19 |
| lucasagomes | but I see that we should try to be awesome, doing on every patch sounds good | 15:19 |
| lucasagomes | I like the idea more now | 15:19 |
| naohirot | good night > all :-) | 15:19 |
| *** naohirot has quit IRC | 15:20 | |
| jroll | rloo: we have to break it sometimes, we should just be careful | 15:20 |
| lucasagomes | rloo, right... yeah it needs to think before, sometimes we can't. Like the arguments. Maybe the new argument should have a default value | 15:20 |
| rloo | jroll: being careful is different that 'must be'. | 15:20 |
| lucasagomes | so it doesn't break others using it | 15:20 |
| rloo | that -> than | 15:21 |
| jroll | rloo: yeah, we should strive for "must be" | 15:21 |
| lucasagomes | it kinda like the kernel idea of not breaking the user space | 15:21 |
| lucasagomes | sometimes they do | 15:21 |
| rloo | ok let's see what others think... | 15:21 |
| lucasagomes | but they try hard to not to | 15:21 |
| lucasagomes | we may want to create a rule like that in ironic as well | 15:21 |
| lucasagomes | don't break CD | 15:21 |
| jroll | ++ | 15:22 |
| rloo | hey, so I've been avoiding sending email to the ML. Do I need to put '[openstack-dev]' in the subject, or just '[ironic]'? | 15:23 |
| jroll | rloo: just [Ironic] | 15:23 |
| jroll | lucasagomes: https://review.openstack.org/#/c/128388/3/ironic/drivers/modules/agent_client.py,cm | 15:24 |
| rloo | jroll: thx. | 15:24 |
| jroll | you're right, want me to update that comment real quick? | 15:24 |
| lucasagomes | jroll, I +2, I mean if u fix I will +2 it again | 15:24 |
| lucasagomes | but I don't think you have to, only for that | 15:24 |
| jroll | yeah ok | 15:24 |
| jroll | /shrug | 15:24 |
| lucasagomes | if happens to u to send a new patch yeah fine | 15:24 |
| openstackgerrit | Jim Rollenhagen proposed openstack/ironic: Support configdrive in agent driver https://review.openstack.org/128388 | 15:25 |
| jroll | there you are sir | 15:25 |
| lucasagomes | hah ok | 15:26 |
| lucasagomes | +2;d | 15:26 |
| jroll | thanks man | 15:26 |
| lucasagomes | yw | 15:26 |
| openstackgerrit | Chris Krelle proposed openstack/python-ironicclient: Update README https://review.openstack.org/134541 | 15:28 |
| NobodyCam | that more along the lines of what we're thinking? | 15:28 |
| jroll | yuriyz: hey, I tested the IPA DIB thing yesterday, it needs more than 2gb RAM to even boot :( want to try to optimize it a bit? | 15:29 |
| yuriyz | hi jroll | 15:30 |
| yuriyz | I tested with 8GB ram | 15:30 |
| yuriyz | I think 4GB is the minimum | 15:30 |
| jroll | yuriyz: I'd like to see if we can run it with 1024mb or even 512mb would be best | 15:31 |
| jroll | if we can get it to run in 512mb we can use it in the gate | 15:31 |
| jroll | the coreos image runs in 1024mb | 15:31 |
| dtantsur | 4GB to boot O_o | 15:33 |
| jroll | nah, just 3GB :P | 15:35 |
| * jroll bbiab | 15:36 | |
| *** jgrimm is now known as zz_jgrimm | 15:36 | |
| NobodyCam | thats a ruff requirment for my laptop testing env | 15:37 |
| yuriyz | I used old server with 8gb in our lab | 15:40 |
| *** ryanpetrello has quit IRC | 15:44 | |
| openstackgerrit | Harshada Mangesh Kakad proposed openstack/ironic: Add serial console feature to seamicro driver https://review.openstack.org/132628 | 15:44 |
| *** ryanpetrello has joined #openstack-ironic | 15:46 | |
| *** dprince has quit IRC | 15:54 | |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic-specs: Discover node properties using new CLI node-discover-properties https://review.openstack.org/100951 | 15:58 |
| *** anderbubble has joined #openstack-ironic | 15:59 | |
| *** Nisha has joined #openstack-ironic | 16:01 | |
| *** andreykurilin_ has joined #openstack-ironic | 16:02 | |
| jroll | yuriyz: I'm glad that it works (or at least boots, I had trouble actually deploying), but with that kind of size it's not very useful, as people could just use the CoreOS image | 16:02 |
| *** ndipanov has quit IRC | 16:03 | |
| jroll | rloo: thanks for that mail | 16:03 |
| rloo | jroll: yw. I debated about whether to list why I didn't agree. I think maybe I should have. The more I think of it, the more I wonder if we even want to try to maintain backwards compatibility. Anyway, let's see what people think. | 16:04 |
| *** Marga_ has joined #openstack-ironic | 16:05 | |
| yuriyz | jroll, CoreOS is optimized for this usecase instead of "generic" distros. What trouble was for deploying? | 16:06 |
| jroll | yuriyz: no, the IPA ramdisk is optimized for space (the container it runs in is a ubuntu image) | 16:07 |
| jroll | idk about deploying | 16:07 |
| jroll | it was late once I got it booting, gave up there as it's not useful to me at this size | 16:07 |
| *** ndipanov has joined #openstack-ironic | 16:07 | |
| NobodyCam | rloo: actually I like the email quite a bit. I love the fact that we strive to maintain backwards compatibility, I am thinking over my reply to the inter-cycle compatibality question | 16:08 |
| NobodyCam | ie we're not set in stone in the actual release cycle's release | 16:08 |
| NobodyCam | if that makes sense | 16:09 |
| rloo | NobodyCam: if folks want to strive for that, that's fine. I'll strive too. But it means yet one more thing for me to think about when reviewing people's patches and I'm worried that it will cause more delays etc | 16:09 |
| *** imtiaz has joined #openstack-ironic | 16:09 | |
| rloo | NobodyCam: anyway, I guess we shouldn't be discussing it here ;) | 16:09 |
| NobodyCam | heheheh | 16:10 |
| jroll | rloo: honestly, anything that won't be backwards compat will likely be in a spec for that reason | 16:10 |
| jroll | but I hear you | 16:10 |
| rloo | jroll: backwards compatible between cycles. no spec may be needed w/i a cycle. | 16:10 |
| yuriyz | jroll, maybe better to use Fedora 21 Atomic after release | 16:10 |
| yuriyz | for DIB | 16:11 |
| *** alexpilotti has joined #openstack-ironic | 16:11 | |
| jroll | rloo: yeah, I take that back | 16:11 |
| devananda | morning, all | 16:11 |
| jroll | heya deva | 16:12 |
| NobodyCam | good morning devananda :) | 16:12 |
| rloo | morning devananda | 16:12 |
| yuriyz | morning devananda | 16:12 |
| rloo | btw, thx for updating the juno release notes devananda! | 16:12 |
| *** imtiaz has quit IRC | 16:13 | |
| * devananda tries to catch up on all the discussion | 16:14 | |
| *** Marga_ has quit IRC | 16:14 | |
| *** andreykurilin_ has quit IRC | 16:14 | |
| *** andreykurilin__ has joined #openstack-ironic | 16:14 | |
| *** Marga_ has joined #openstack-ironic | 16:14 | |
| *** anderbubble has quit IRC | 16:15 | |
| *** blinky_ghost_ has joined #openstack-ironic | 16:17 | |
| *** jiangfei has quit IRC | 16:19 | |
| *** jiangfei has joined #openstack-ironic | 16:19 | |
| *** Marga_ has quit IRC | 16:19 | |
| * NobodyCam wants to change "black friday" to "the two weeks of spam" | 16:19 | |
| *** dlaube has joined #openstack-ironic | 16:20 | |
| *** Marga_ has joined #openstack-ironic | 16:20 | |
| openstackgerrit | Jim Rollenhagen proposed openstack/python-ironicclient: Add maintenance_reason to node-show output https://review.openstack.org/135999 | 16:22 |
| jroll | ^ fixed tests | 16:23 |
| * jroll bbl | 16:23 | |
| *** Nisha has quit IRC | 16:26 | |
| *** foexle has quit IRC | 16:26 | |
| *** zz_jgrimm is now known as jgrimm | 16:29 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic-specs: Support for non-glance image references https://review.openstack.org/135276 | 16:36 |
| blinky_ghost_ | Hi all, can anybody help me with issue: I'm using tripleo with ironic to boot a node Dell R320 and I see this error on the screen: "ipmitool - can't find /dev/ipmi0 or /dev/ipmidev/0" then the node shutdowns itself. thanks | 16:37 |
| lucasagomes | devananda, morning | 16:37 |
| NobodyCam | blinky_ghost_: where are you seeing that error? | 16:38 |
| blinky_ghost_ | NobodyCam: I'm my overcloud baremetal node, on the screen. | 16:39 |
| lucasagomes | blinky_ghost_, ok so you deployed the node with tripleo + ironic | 16:41 |
| devananda | rloo: backwards compat w/in master w/in a cycle -- yes, we should very much care about this, within reason | 16:41 |
| lucasagomes | blinky_ghost_, and in the node's shell you try to use ipmitool and u get that error? | 16:41 |
| NobodyCam | humm, is the overcloud attemptng to provision a baremetal node or VM? | 16:41 |
| *** ChuckC has quit IRC | 16:41 | |
| lucasagomes | blinky_ghost_, it's not in the Ironic conductor node right? | 16:41 |
| blinky_ghost_ | nobodycam: baremetal | 16:42 |
| rloo | devananda: what does 'within reason' mean then? | 16:42 |
| lucasagomes | there's a ML thread about the backwards compat now, I added my 2 cents there | 16:42 |
| lucasagomes | (thanks rloo for starting it) | 16:42 |
| blinky_ghost_ | lucasgomes: the node boots kernel autodiscovery ramdisk gives that message and shutdowns | 16:42 |
| lucasagomes | blinky_ghost_, right, which ramdisk is that? You created it | 16:43 |
| lucasagomes | it looks like the ramdisk doesn't have the required ipmi kernel modules loaded | 16:43 |
| lucasagomes | so u can issue ipmitool commands in-band | 16:43 |
| NobodyCam | blinky_ghost_: AIU the overcloud will provision VM's not baremetal, Baremetal is provisioned by seed and undercloud | 16:43 |
| lucasagomes | You created it? (was a question) | 16:44 |
| devananda | rloo: we should be testing (and i believe adam_g is working on it) compat in two ways: last release to tip; tip-1 to tip; | 16:44 |
| blinky_ghost_ | NobodyCam: I'm using undercloud to provision overcloud, using instack | 16:44 |
| blinky_ghost_ | lucasgomes: any good howto to include this in the on the ramdisk? | 16:45 |
| dtantsur | blinky_ghost_, you'd better mention that it's about discovery ramdisk | 16:46 |
| lucasagomes | blinky_ghost_, well like copying any kernel module to a ramdisk, might have a couple of tutorials on google | 16:46 |
| rloo | devananda: the tip-1 to tip tests will be interesting. | 16:46 |
| blinky_ghost_ | lucasgomes: ok thanks | 16:47 |
| dtantsur | lucasagomes, it should work as of now too with `modprobe ipmi_devintf` workaround, I have no idea why it does not | 16:47 |
| lucasagomes | u just need to copy the right ipmi modules from the /usr/lib/modules/<kernel version> to ur ramdisk | 16:47 |
| lucasagomes | dtantsur, yup, if the module is there | 16:47 |
| *** pensu has joined #openstack-ironic | 16:47 | |
| blinky_ghost_ | dtantsur: does it help if I enable debug in ironic-discovery? | 16:48 |
| dtantsur | blinky_ghost_, no, it won't help. the ramdisk is quite an isolated thing... | 16:48 |
| *** imtiaz has joined #openstack-ironic | 16:49 | |
| *** imtiaz has quit IRC | 16:49 | |
| *** imtiaz has joined #openstack-ironic | 16:49 | |
| blinky_ghost_ | dtantsur: ok, so it shoud work as you're saying that module is there...maybe is not beeing loaded. It has to run modrpobe module? | 16:49 |
| dtantsur | yeah... maybe it should modprobe more modules... i'm trying to google it | 16:50 |
| dtantsur | blinky_ghost_, any output on the screen mentioning modprobe or ipmi_devintf? | 16:51 |
| * devananda is less distracted now, still catching up on backlog | 16:51 | |
| lucasagomes | blinky_ghost_, dtantsur http://serverfault.com/questions/480371/ipmitool-cant-find-dev-ipmi0-or-dev-ipmidev-0 | 16:52 |
| dtantsur | ipmi_si... maybe it is missing | 16:52 |
| blinky_ghost_ | dtantsur: maybe I can include the modprobe commands in the ramdisk | 16:54 |
| dtantsur | blinky_ghost_, what if you try to cope line 40 of the ramdisk and make it load ipmi_si as well? | 16:55 |
| *** anderbubble has joined #openstack-ironic | 16:56 | |
| *** marcoemorais has joined #openstack-ironic | 16:57 | |
| lucasagomes | dtantsur, u can create a /etc/modules.conf | 16:57 |
| blinky_ghost_ | dtantsur: what's the file? i have discovery-ramdisk.initramfs and discovery-ramdisk.kernel but they're binary | 16:57 |
| lucasagomes | the modules listed there will be loaded in boot | 16:57 |
| lucasagomes | blinky_ghost_, initramfs, you can extract it | 16:57 |
| *** jiangfei has quit IRC | 16:58 | |
| lucasagomes | blinky_ghost_, http://nst.sourceforge.net/nst/docs/faq/ch10s13.html | 16:58 |
| *** jiangfei has joined #openstack-ironic | 16:58 | |
| openstackgerrit | Josh Gachnang proposed openstack/ironic-python-agent: Add command metrics to IPA API https://review.openstack.org/119981 | 16:58 |
| *** ifarkas has quit IRC | 16:59 | |
| dtantsur | blinky_ghost_, well yes, you got these with instack... maybe it's better to me to try patch instack and then you retry | 16:59 |
| devananda | rloo: so I may be missing the context -- what's the concrete thing that is or is not backwards compatible? | 16:59 |
| lucasagomes | devananda, it was about renaming the @passthru decorator to @node_passthru (because we also have a @driver_passthru, @node_passthru would be more explicit) | 17:00 |
| devananda | lucasagomes: gotcha. the decorator, not the REST API | 17:00 |
| blinky_ghost_ | dtanstur: OK, thanks. I downloaded the images from here: https://repos.fedorapeople.org/repos/openstack-m/tripleo-images-rdo-juno/ | 17:01 |
| lucasagomes | and jroll was pointing out that if we rename it, even merged few days ago, we should still have a backward compat module, because drivers out of tree might be using it | 17:01 |
| lucasagomes | devananda, yes | 17:01 |
| lucasagomes | even if* | 17:01 |
| *** andreykurilin__ has quit IRC | 17:01 | |
| lucasagomes | (my english is going down, damn) | 17:01 |
| devananda | "a few days ago" -- i'm fine with just fixing it | 17:01 |
| devananda | a) it's internal b) it's just a few days | 17:01 |
| lucasagomes | well, it's not super internal, potentially there are drivers out of tree that might be using it | 17:02 |
| lucasagomes | and the fix would be trivial, just an alias passthru = node_passthru | 17:02 |
| lucasagomes | so I'm +1 to make it backward compat | 17:02 |
| devananda | if someone downstream is depending on an internal python decorator that's only been in trunk for a few days, they can handle the change. | 17:02 |
| lucasagomes | they could, but would that make them happy? | 17:03 |
| lucasagomes | I mean we are potentially breaking their workflow | 17:03 |
| lucasagomes | if we can avoid it, adding 1 extra line of code | 17:03 |
| lucasagomes | I would do it | 17:03 |
| devananda | yup, then we should | 17:03 |
| lucasagomes | +1 | 17:05 |
| lucasagomes | but anyway, that example was just the trigger for the discussion about whether have backwards compat on every commit or not | 17:05 |
| lucasagomes | which I think we should at least try (try hard) to keep the compat | 17:06 |
| lucasagomes | on every commit | 17:06 |
| *** anderbubble has quit IRC | 17:06 | |
| dtantsur | slagle, hi https://github.com/agroup/instack-undercloud/pull/83 should fix the problem blinky_ghost_ encountered above | 17:09 |
| NobodyCam | added my two cents to the ML chain | 17:13 |
| blinky_ghost_ | dtantsur: thanks :) I'll test it out | 17:13 |
| *** anderbubble has joined #openstack-ironic | 17:14 | |
| *** ndipanov has quit IRC | 17:14 | |
| *** sambetts has quit IRC | 17:15 | |
| blinky_ghost_ | dtantsur: I have to create a new image after aplying the patch? | 17:15 |
| dtantsur | blinky_ghost_, yes | 17:16 |
| *** andreykurilin_ has joined #openstack-ironic | 17:16 | |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic-specs: Automate UEFI-BIOS Iso Creation https://review.openstack.org/134022 | 17:18 |
| blinky_ghost_ | dtantsur: ramdisk-image-create fedora deploy-ironic -o my-deploy-ramdisk ? | 17:19 |
| *** ChuckC has joined #openstack-ironic | 17:19 | |
| rloo | lucasagomes: thx for explaining. Although it might be more like 1-2 weeks ago (was it sambett's change that had the @passthru?) | 17:19 |
| lucasagomes | rloo, oh yeah | 17:19 |
| dtantsur | blinky_ghost_, $ ramdisk-image-create fedora discovery-ironic -o my-deploy-ramdisk | 17:19 |
| lucasagomes | yeah I think it may be a bit older | 17:19 |
| * devananda replies to the thread | 17:20 | |
| lucasagomes | rloo, it's more a philosophical approach, but I think we should try our best | 17:20 |
| blinky_ghost_ | dtantsur: ok thanks testing | 17:20 |
| rloo | lucasagomes: I'm fine going with the majority. I just know that I haven't been reviewing things with this in mind. | 17:20 |
| lucasagomes | rloo, yeah, myself included. I think that after this conversation something clicked in my mind | 17:21 |
| lucasagomes | I will pay more attention too | 17:21 |
| lucasagomes | it's really important | 17:21 |
| rloo | ah, just read dtantsur's reply, I like his ;) ok, I'll be quiet now ;) | 17:22 |
| dtantsur | :) | 17:22 |
| devananda | rloo: thanks for bringing it up. it was less important prior to graduation; now we have many more people counting on it | 17:22 |
| *** anderbubble has quit IRC | 17:23 | |
| devananda | lucasagomes: also s/@passthru/@node_passthru/ is, IIUC, just a name change. it's not a functionality change | 17:24 |
| *** dtantsur is now known as dtantsur|afk | 17:24 | |
| *** Viswanath has joined #openstack-ironic | 17:24 | |
| rloo | devananda: does it make a diff is it is a functionality vs name change? the name change (w/o aliasing it) means that an out of tree driver won't work | 17:25 |
| lucasagomes | devananda, yup yeah it's just a name change | 17:25 |
| lucasagomes | devananda, the tihng is that if we rename it, we still break potential users | 17:25 |
| devananda | rloo: if it's "just a name change" then, to be frank, it looks silly if we break anyone over it | 17:25 |
| devananda | from a user/developer's point of view, how aggravating would that be? | 17:26 |
| devananda | the relative importance of the change is not irrelevant here | 17:26 |
| devananda | "oh, they found a massive bug in $newfeature and had to do a breaking change to fix it" --> I could deal with that | 17:26 |
| devananda | "they changed their mind about the name of @thing and broke me? wtf?" | 17:27 |
| lucasagomes | indeed | 17:27 |
| lucasagomes | yeah sometimes I think we are going to see some things like, a critical bug that we will have to break it | 17:27 |
| devananda | whether or not we /think/ someone should or should not depend on trunk -- it's the nature of continuous deployment that they /will/ | 17:27 |
| *** Viswanath has quit IRC | 17:27 | |
| rloo | devananda: right, between releases. But w/i development of a release, should people expect that things will 'still' work? | 17:27 |
| lucasagomes | also if we try hard to not break things on daily basis, it will much simpler to not break between releases | 17:28 |
| devananda | lucasagomes: ++ | 17:28 |
| *** ndipanov has joined #openstack-ironic | 17:28 | |
| devananda | rloo: whether we think they should or not, I can tell you that there are folks who operate that way on a daily basis w.r.t. other projects (Nova, Swift, Glance, Keystone ... all come to mind) | 17:29 |
| rloo | devananda: hmm. ok. (Just read your email). Although I'm not sure I agree with lucasagomes (I don't see the logic there) | 17:29 |
| *** jiangfei has quit IRC | 17:29 | |
| devananda | rloo: I happen to agree that any project which calls itself mature enough for continuous deployment (and by virtue of being Integrated with OpenStack, we've declared that we think Ironic is) should strive for that | 17:30 |
| devananda | we'll fail some times | 17:30 |
| *** jiangfei has joined #openstack-ironic | 17:30 | |
| *** Nisha has joined #openstack-ironic | 17:31 | |
| lucasagomes | rloo, right. Does it sounds too bureaucratic ? | 17:31 |
| lucasagomes | or even hypocrite since, I'm sure that at one point we are going to break it | 17:31 |
| *** ndipanov is now known as ndipanov_gone | 17:31 | |
| rloo | lucasagomes: no, not bureaucratic. We may accidentally break it, but at least, I know that we don't want something broken ;) | 17:32 |
| devananda | lucasagomes: in your email, second paragraph -- s/there may be/there are/ | 17:32 |
| lucasagomes | rloo, sure... Yeah it can happen and will happen | 17:32 |
| devananda | lucasagomes: just ask J* what version of Ironic they're running | 17:32 |
| rloo | lucasagomes: i think this means we really need to pay more attention to any proposed changes to any of the APIs. | 17:32 |
| lucasagomes | devananda, right... +1 yeah so it's even more incentive to not try to break the guys :D | 17:33 |
| lucasagomes | cause we are going to meet them on the next summit :p | 17:33 |
| devananda | rloo: there's a reason the spec process calls out each API, and then a separate section for upgrades and backwards compat :) | 17:33 |
| rloo | devananda: yes, but I was thinking between cycles, not w/i a cycle. | 17:33 |
| lucasagomes | rloo, yup yeah that's pretty much it. Let's pay more attention | 17:33 |
| devananda | rloo: ahhh. ok. so yea, this is a great discussion to clarify that | 17:34 |
| devananda | also, adam_g, are you around? | 17:34 |
| NobodyCam | lucasagomes: +1^398643 | 17:34 |
| adam_g | devananda, heya | 17:34 |
| rloo | devananda: cuz sometimes the spec proposes something, the code is written and then we decide at some point we don't like the proposed API. I don't recall if we've changed before something has landed, or after. | 17:34 |
| *** Haomeng has joined #openstack-ironic | 17:35 | |
| devananda | adam_g: hi! you might want to skim scrollback / read the thread ruby started... it's all about upgrades :) | 17:35 |
| adam_g | devananda, yeah, catching up now :) | 17:35 |
| adam_g | just got online | 17:35 |
| *** Haomeng|2 has quit IRC | 17:35 | |
| devananda | rloo: yup. sometimes the spec isn't clear enough -- or sometimes we see the code and realize something we didn't think of when reviewing the spec. that happens | 17:36 |
| devananda | rloo: in that case, we should go revise the spec, not land a crappy API | 17:36 |
| devananda | though I actually don't recall that happening at all last cycle ... ? | 17:36 |
| rloo | devananda: right. But it looks like we need to be more diligent about making sure a 'not-so-great' API doesn't land first before noticing it. | 17:37 |
| devananda | oh, also a note for lucasagomes and GheRivero -- see Thierry's email about "a true cross-project meeting" | 17:37 |
| devananda | rloo: ++ | 17:37 |
| rloo | devananda: to be honest, I don't know if it happened or not last cycle. I do know that I have mentioned/asked for changes to APIs. I just don't recall now if it was via the spec, a patch, or after a patch has landed. | 17:38 |
| lucasagomes | devananda, oh it's on the openstack-dev? or -tc? | 17:38 |
| devananda | rloo: fwiw, I don't recall seeing any breaking changes to APIs land last cycle, so I think we're doing OK :) | 17:39 |
| * lucasagomes dig into his mails | 17:39 | |
| devananda | lucasagomes: dev | 17:39 |
| devananda | lucasagomes: heading [all] | 17:39 |
| lucasagomes | ack | 17:39 |
| devananda | actually, adam_g, you might want to look at that email from thierry as well, since you're sort of the unofficial QA liaison (or is it official now?) | 17:40 |
| rloo | devananda: eg, in this particular case, I didn't see the decorator name land, but I did mention it in some subsequent patch that used it, and then brought it up again. The decorator isn't that big a deal, but here we are ;) | 17:40 |
| devananda | rloo: indeed, the decorator change has a trivialy simple backwards-compatible-implementation -- but the larger point is well timed to align everyone onto the same expectation | 17:42 |
| rloo | devananda: yup! | 17:42 |
| *** romcheg has quit IRC | 17:43 | |
| *** romcheg has joined #openstack-ironic | 17:43 | |
| *** romcheg has quit IRC | 17:44 | |
| lucasagomes | +1 for adam_g be the QA liaison :) (if he wants ofc) | 17:44 |
| *** anderbubble has joined #openstack-ironic | 17:49 | |
| adam_g | sure! | 17:49 |
| *** romcheg has joined #openstack-ironic | 17:55 | |
| *** derekh has quit IRC | 17:58 | |
| *** achanda has joined #openstack-ironic | 17:59 | |
| lucasagomes | I will call it a day, I'm in the office today so gotta get the train back | 18:01 |
| lucasagomes | have a good night everyone | 18:01 |
| *** harlowja_away is now known as harlowja | 18:02 | |
| rloo | night lucasagomes | 18:02 |
| *** dprince has joined #openstack-ironic | 18:03 | |
| jroll | night lucasagomes | 18:04 |
| NobodyCam | night lucasagomes | 18:04 |
| *** spandhe has joined #openstack-ironic | 18:05 | |
| adam_g | devananda, so re: grenade, we'll be testing last release to tip. tip-1 to tip is the sideways stuff that is really only used to test the nova-bm migration on juno. we can probably leverage it to do continued sideways testing of master if we realy wanted to, but im not sure that would help the issue of breaking out of tree drivers, or even in-tree, non-default drivers | 18:05 |
| *** linggao has quit IRC | 18:05 | |
| *** penick has joined #openstack-ironic | 18:07 | |
| *** lucasagomes has quit IRC | 18:08 | |
| *** marcoemorais has quit IRC | 18:08 | |
| *** marcoemorais has joined #openstack-ironic | 18:08 | |
| *** krtaylor has quit IRC | 18:13 | |
| *** krtaylor has joined #openstack-ironic | 18:16 | |
| *** ryanpetrello_ has joined #openstack-ironic | 18:22 | |
| *** subscope has joined #openstack-ironic | 18:23 | |
| openstackgerrit | Chris Krelle proposed openstack/python-ironicclient: Update README https://review.openstack.org/134541 | 18:23 |
| *** MattMan has quit IRC | 18:24 | |
| *** ryanpetrello has quit IRC | 18:25 | |
| *** ryanpetrello_ is now known as ryanpetrello | 18:25 | |
| NobodyCam | brb quick walkies | 18:26 |
| *** marcoemorais has quit IRC | 18:36 | |
| *** marcoemorais has joined #openstack-ironic | 18:36 | |
| *** marcoemorais has quit IRC | 18:40 | |
| *** marcoemorais has joined #openstack-ironic | 18:41 | |
| *** imtiaz has quit IRC | 18:41 | |
| *** rushiagr is now known as rushiagr_away | 18:45 | |
| *** dprince has quit IRC | 18:45 | |
| * NobodyCam is almost back | 18:46 | |
| *** andreykurilin_ has quit IRC | 18:52 | |
| *** Marga_ has quit IRC | 18:52 | |
| *** Marga_ has joined #openstack-ironic | 18:52 | |
| *** Viswanath has joined #openstack-ironic | 19:00 | |
| *** igordcard has quit IRC | 19:04 | |
| *** marcoemorais has quit IRC | 19:04 | |
| *** marcoemorais has joined #openstack-ironic | 19:05 | |
| *** eghobo has joined #openstack-ironic | 19:06 | |
| *** Viswanath has quit IRC | 19:07 | |
| *** imtiaz has joined #openstack-ironic | 19:10 | |
| *** eghobo has quit IRC | 19:13 | |
| *** pelix has quit IRC | 19:15 | |
| PaulCzar | where's the ironic standalone stuff at? having a hard time finding docs and info about it | 19:16 |
| NobodyCam | docs what docs | 19:18 |
| NobodyCam | lol there really are not any atm | 19:18 |
| jroll | yeah, we don't have docs on that stuff yet | 19:18 |
| jroll | it's something we'd like to do but is not quite there | 19:18 |
| jroll | like, you'll still need glance | 19:18 |
| openstackgerrit | Merged openstack/ironic: Add driver_validate() https://review.openstack.org/134184 | 19:24 |
| *** eghobo has joined #openstack-ironic | 19:26 | |
| devananda | PaulCzar: we don't formally support stand-alone Ironic today. It's tedious to set up and use that way, and I suspect that only a few of us have actually done it in a lab yet | 19:27 |
| devananda | PaulCzar: a great deal of logic around deployments exists within the nova.virt.ironic driver code | 19:28 |
| devananda | *only within ... today | 19:28 |
| *** rwsu has quit IRC | 19:32 | |
| *** marcoemorais has quit IRC | 19:33 | |
| *** blinky_ghost_ has quit IRC | 19:33 | |
| *** rwsu has joined #openstack-ironic | 19:33 | |
| *** marcoemorais has joined #openstack-ironic | 19:33 | |
| *** marcoemorais has quit IRC | 19:33 | |
| *** marcoemorais has joined #openstack-ironic | 19:33 | |
| NobodyCam | added another name to the list :) | 19:36 |
| adam_g | devananda, shall i throw my name down on the QA list here? https://wiki.openstack.org/wiki/CrossProjectLiaisons | 19:41 |
| devananda | adam_g: ++ | 19:41 |
| devananda | adam_g: are you still doing stable branch stuff? | 19:41 |
| adam_g | devananda, yeah, more so with release management and keeping branches in good health | 19:42 |
| *** krtaylor has quit IRC | 19:45 | |
| *** imtiaz has quit IRC | 19:47 | |
| devananda | adam_g: awesome. you're welcome to add your name to that CPL list as well | 19:48 |
| devananda | adam_g: also, I now have +2/+A rights on stable, so between the two of us we can actually fix broken things :) | 19:48 |
| devananda | adam_g: don't hesitate to bug me if there's a patch on stable/juno that needs to land -- I may not look all that often otherwise | 19:48 |
| adam_g | devananda, sure. we might want to get dtantsur|afk involved there--he's up on all the new bugs and could probably help identify potential backports quicker than me | 19:49 |
| devananda | sounds good | 19:49 |
| *** Marga_ has quit IRC | 19:54 | |
| *** krtaylor has joined #openstack-ironic | 19:58 | |
| *** penick has quit IRC | 19:59 | |
| *** marcoemorais has quit IRC | 19:59 | |
| *** linggao has joined #openstack-ironic | 20:00 | |
| *** pensu has quit IRC | 20:00 | |
| NobodyCam | can get folks thoughts on Ironic creating deployment discs (question is comming from a spec to create an iso, but I expect it could be extended to create a deploy K&R too) | 20:02 |
| devananda | NobodyCam: I've been fairly strongly against Ironic creating long-lived images in the past | 20:07 |
| devananda | that said, the ilo driver *does* build images and upload to swift | 20:07 |
| NobodyCam | and we are getting in to the config drive biz too | 20:08 |
| devananda | right - but taht's metadata | 20:08 |
| devananda | for config drive, we will accept a reference to a gzip'd ISO, pull from swift, write to disk -- that's well within what Ironic alraedy did | 20:08 |
| devananda | the difference for iLO is that it creatse a vfat image on the fly, holding potentially sensitive information, saves to swift with a short TTL, and then attaches that to the node | 20:09 |
| devananda | it's the only case yet AFAIK where Ironic is /building/ the image (rather than consuming it) | 20:09 |
| devananda | so I guess my question is, in the spec you're referencing, what is the function of the ISO being created? | 20:09 |
| NobodyCam | its apears to be a deployment iso ... grabbing link | 20:10 |
| NobodyCam | https://review.openstack.org/#/c/134022/3/specs/kilo/automate-uefi-bios-iso-creation.rst | 20:10 |
| NobodyCam | i ask because at first read I am opposed to ironic doing such things, but I also don't want to be the "stick in the mud" that stands agenst the tide of opinion | 20:14 |
| * NobodyCam thinks he could have gotten a few more medifors in that last statment | 20:15 | |
| * devananda wants to eat medifors now | 20:21 | |
| devananda | I'm drafting the midcycle announcement email ... | 20:22 |
| devananda | jroll, JayF: IIRC ya'll said ya'll couldn't make it to EU, but had some ideas on improving remote participation | 20:22 |
| Nisha | NobodyCam: devananda hi. Yes thats the spec i raised for just automating boot iso which is required for iscsi-ilo driver and is already done fo rbios mode | 20:23 |
| devananda | Nisha: is there a difference between bios and uefi mode, with regard to the ISO or vfat image created? | 20:24 |
| Nisha | devananda: the current ISO creation is just using syslinux which is not capable of uefi booting. For UEFI , we need to use EFI bootloader | 20:25 |
| Nisha | In juno we tried to do this using ELILO | 20:26 |
| NobodyCam | Nisha: ahh I think I missed the context that this was improving what we already do with bios mode ilo deploys | 20:26 |
| Nisha | devananda: but it didnt pass the tests, hence backed out the code at the last moment | 20:26 |
| Nisha | devananda: yes | 20:26 |
| Nisha | NobodyCam: devananda i am aware that in Juno we were against auto creating deploy ISOs. Since the same piece of code can be used to even create deploy isos for iscsi-ilo driver, is it worth to extend that support for deploy iso? | 20:29 |
| jroll | devananda: we said we likely cannot make it, there's a small chance, and JayF is happy to help with remote participation | 20:30 |
| devananda | Nisha: I don't follow - the deploy ISO contains significantly more than just the bootloader | 20:31 |
| devananda | Nisha: it also contains, for example, ironic-python-agent | 20:31 |
| devananda | jroll: I like it when I remember things | 20:31 |
| jroll | persistent memory is nice to have :) | 20:32 |
| *** kfox1111 has joined #openstack-ironic | 20:32 | |
| Nisha | devananda: Yes it contains the ramdisk and kernel along with boot loader | 20:32 |
| kfox1111 | question. The documentation says to create a net, but it doesn't say to create a subnet. Where do the ip's come from? | 20:32 |
| devananda | jroll: I'm assuming "help with remote participation" extends to non-rackers who might not be able to attend as well | 20:32 |
| devananda | kfox1111: neutron | 20:32 |
| kfox1111 | a subnet is not mentioned in the docs. Are they incorrect, and you do need to add a subnet? | 20:33 |
| *** mrda_away is now known as mrda | 20:33 | |
| devananda | Nisha: so that is the part that strikes me as an unnecessary scope expansion | 20:33 |
| mrda | morning Ironic | 20:33 |
| Nisha | devananda: i havent explored for automating deploy iso in ironic, but if kernel and ramdisk are inputted to the function it ca create the deploy iso | 20:33 |
| devananda | Nisha: "can" != "should" | 20:34 |
| jroll | devananda: yes, ofc | 20:34 |
| jroll | morning mrda ) | 20:34 |
| jroll | :) even | 20:34 |
| devananda | Nisha: is that deploy ISO going to be unique to the node? unique to the instance? or common across many nodes and instances? | 20:34 |
| devananda | g'morning, mrda! | 20:34 |
| Nisha | it will unique to the node as we do for boot iso today, because then it will be created on the fly | 20:35 |
| devananda | Nisha: why is it unique to the node? | 20:35 |
| devananda | that's what I don't understand | 20:35 |
| devananda | deploy images are common artefacts - they should work on any similar hardware, not just on a specific node. Unless I'm misunderstanding something about UEFI? | 20:36 |
| jroll | what is special about UEFI where it must be unique to the node? | 20:36 |
| Nisha | the same iso can be used for deploying many nodes if it is available in swift forever, but in boot iso we tend to delete the iso as the node is deactivated | 20:37 |
| jroll | I'm looking at "problem description" here and I don't understand the problem https://review.openstack.org/#/c/134022/3/specs/kilo/automate-uefi-bios-iso-creation.rst | 20:37 |
| jroll | why do we need to put the deploy ISO in swift? | 20:37 |
| devananda | 20:37:07 < Nisha> the same iso can be used for deploying many nodes if it is available <<<< then it should be stored in Glance | 20:37 |
| Nisha | jroll: the boot iso is specific for the deploy image | 20:37 |
| jroll | why does ironic need to generate it? | 20:37 |
| jroll | the boot iso, yes | 20:37 |
| jroll | ok, which is what the spec is for | 20:38 |
| Nisha | devananda: yes...thats why i just asked if it is worth to extend the support for deploy iso | 20:38 |
| jroll | why are we talking about deploy ISOs? | 20:38 |
| jroll | oh, no | 20:38 |
| Nisha | i havent proposed it yet | 20:38 |
| Nisha | :) | 20:38 |
| jroll | ok :) | 20:38 |
| * jroll will read better next time | 20:38 | |
| devananda | Nisha: ah! now I understand your question :) | 20:38 |
| Nisha | devananda: thanks | 20:38 |
| Nisha | devananda: since today i am able to catch you (my night time) , request your review for https://review.openstack.org/#/c/100951/42 also (node-properties discovery) | 20:39 |
| *** tatyana has joined #openstack-ironic | 20:40 | |
| devananda | Nisha: ack. reading .. | 20:40 |
| devananda | Nisha: have you been following the state machine change proposal? | 20:41 |
| Nisha | devananda: yes | 20:41 |
| Nisha | discovery states are still in discussion | 20:41 |
| Nisha | i have based my spec on that spec as of now | 20:42 |
| Nisha | jroll: two days back i asked for agent_ilo driver to be supported for uefi....Jayf and you suggested taht the current mechanism for agent_ilo should work even for uefi boot if image is hybrid one and you provided me the link to one such image too | 20:42 |
| Nisha | jroll: but the deploy iso created for this doesnt work for uefi :( | 20:43 |
| Nisha | deploy iso is created using ironic-python-agent | 20:43 |
| jroll | Nisha: that image was an example of an image with GPT, I don't know if it works with UEFI | 20:44 |
| Nisha | for agent-ilo driver | 20:44 |
| jroll | that image was just to show that IPA can deploy whole-disk images with GPT | 20:44 |
| Nisha | jroll thats correct , but the deploy iso itself is not working | 20:44 |
| Nisha | :( | 20:44 |
| jroll | which deploy iso? did you build it yourself or? | 20:45 |
| Nisha | image is booted in second boot...the deploy iso should also be uefi capable | 20:45 |
| jroll | I've never used that, I forget who wrote that code | 20:45 |
| *** tatyana has quit IRC | 20:45 | |
| Nisha | yes | 20:45 |
| jroll | yeah, no idea if that deploy iso supports uefi | 20:45 |
| Nisha | jroll http://docs.openstack.org/developer/ironic/drivers/ilo.html#drivers , the Configuring and Enabling the driver section of the agent_ilo | 20:46 |
| Nisha | means IPA needs to be enhanced for supporting uefi | 20:47 |
| jroll | Nisha: yes, I know this exists | 20:47 |
| Nisha | atleast the scripts which are used to create the deploy iso | 20:47 |
| jroll | can you please put up a patch to fix it? | 20:47 |
| jroll | I don't know enough about uefi or building ISOs to quickly fix it | 20:48 |
| jroll | or at least file a bug that someone (ramesh?) can be pointed at | 20:48 |
| Nisha | jroll: I was planning to propose the spec for enabling agent_ilo for uefi | 20:48 |
| Nisha | can it be taken care under that? | 20:48 |
| jroll | oh, is there other work to be done for that? | 20:49 |
| jroll | then yes, let's do it under that spec | 20:49 |
| jroll | idk if we need a spec for that work, but I guess it doesn't hurt | 20:49 |
| Nisha | jroll: yes, the boot modes need to switch as suggested in the flavor by the user | 20:49 |
| jroll | right | 20:49 |
| Nisha | ok. I will propose one | 20:49 |
| *** andreykurilin_ has joined #openstack-ironic | 20:50 | |
| Nisha | devananda: i have a query for nova specs (specs required for pushing flavor keys to ironic), whom shall i contact there??? | 20:51 |
| Nisha | devananda: reference https://review.openstack.org/136104 | 20:51 |
| devananda | Nisha: comments posted | 20:51 |
| Nisha | devananda: thanks. WIll just read and respond | 20:51 |
| Nisha | devananda: i even posted this one https://review.openstack.org/133534 to enhance ComputeCapability filter to parse mulitple values for a key | 20:52 |
| devananda | Nisha: does that require a change in Nova outside of the "nova.virt.ironic.*" tree? | 20:53 |
| Nisha | The first one is in this tree only | 20:53 |
| Nisha | the second one is in filters | 20:54 |
| Nisha | The ComputeCapabilities filter will be in nova/scheduler/filters/compute_capabilities_filter.py | 20:55 |
| Nisha | The other one will be in nova/virt/ironic/patcher.py | 20:55 |
| devananda | Nisha: you should raise it in the Nova IRC channel and/or weekly meeting. they follow a similar blueprint / spec process to Ironic, but have a *lot* more contributors and the review team is under a lot more strain ... | 20:55 |
| devananda | Nisha: I suspect changes purely inside of nova.virt.ironic.* will be much easier to get sign-off on, and please add me to the review list | 20:56 |
| devananda | a change in nova/scheduler/filters/* is going to take more agreement, since it will affect all users of nova | 20:56 |
| Nisha | hmmmm, ok for ComputeCapabilities i wanted your/core team opinion here since it is related to hardware capabilities | 20:56 |
| Nisha | It is just an enhancement to what that filter do today | 20:57 |
| devananda | Nisha: do you know how to add a reviewer to a review? | 20:57 |
| Nisha | devananda: no, i havent done so far | 20:57 |
| devananda | ok - for example, on https://review.openstack.org/#/c/133534/2 , under the section titled "Reviewer" there is a text entry field | 20:57 |
| devananda | this uses autocomplete to search for gerrit users | 20:58 |
| devananda | my id is "devananda" | 20:58 |
| Nisha | Ok | 20:58 |
| Nisha | devananda: done | 20:58 |
| devananda | please add me to nova spec reviews that affect Ironic, like these two -- I would like to review them, but probably won't notice them unless someone adds me (or pings me on IRC like this :) ) | 20:59 |
| Nisha | devananda: :) ok i will add you in another also | 20:59 |
| devananda | Nisha: I do not see it yet. did you click the "Add Reviewer" button? | 20:59 |
| Nisha | Yes i added you in both | 21:00 |
| devananda | ahh - see it now. thanks! | 21:01 |
| Nisha | for hardware capabilities how do we want to go about? there are two possible approaches: | 21:01 |
| Nisha | 1. have a capability which can have one or more values, like Supported_Boot_Modes=['uefi','bios'] | 21:02 |
| Nisha | if we do this, it requires changes in ComputeCapability Filter to support multiple values for a key | 21:03 |
| Nisha | which i have proposed above. | 21:03 |
| Nisha | 2. Have boolean values per capability | 21:03 |
| Nisha | example, uefi_mode=True, bios_mode=True | 21:03 |
| Nisha | This approach doesnt require any changes in ComputeCapabilities Filter | 21:04 |
| Nisha | but it means we need to identify all possible values for a capability | 21:04 |
| Nisha | which may vary hardware to hardware | 21:04 |
| Nisha | what shall be ironic recommendation | 21:05 |
| Nisha | ? | 21:05 |
| NobodyCam | i'd pick #1 if it where up to /me | 21:06 |
| jroll | I think we need to discuss this in the capabilities spec | 21:07 |
| jroll | before we push specs into nova | 21:07 |
| NobodyCam | jroll: + | 21:07 |
| Nisha | jroll: yes | 21:07 |
| jroll | so uh | 21:07 |
| jroll | someone should write that spec | 21:07 |
| * jroll runs away | 21:07 | |
| kfox1111 | Is there an example of a ironic node-create with pxe_ipmitool somewhere? | 21:07 |
| Nisha | but since nova specs take time to get approval, so just suggested an alternative | 21:07 |
| Nisha | jroll: already proposed to Nova for ComputeCapabilities Filter | 21:08 |
| NobodyCam | kfox1111: see last step of http://docs.openstack.org/developer/ironic/deploy/install-guide.html#flavor-creation | 21:08 |
| Nisha | https://review.openstack.org/133534 | 21:08 |
| jroll | Nisha: I know, I don't want to discuss the same thing in two different specs | 21:08 |
| kfox1111 | nobodycam: ah, there we go. I missed that part. Thanks. | 21:09 |
| NobodyCam | :) | 21:09 |
| Nisha | jroll: wan-yen already gave comments in https://review.openstack.org/131272 | 21:09 |
| Nisha | which is in ironic | 21:09 |
| Nisha | she suggested the workaround of boolean values | 21:10 |
| jroll | Nisha: I see that | 21:10 |
| jroll | others should comment too | 21:10 |
| jroll | and we shold finish the spec | 21:10 |
| Nisha | jroll: :) | 21:10 |
| jroll | but I don't want to discuss this in two places | 21:10 |
| jroll | that's silly | 21:10 |
| Nisha | we also prefer the first approach | 21:10 |
| jroll | we should WIP the nova specs until we land the ironic specs IMO | 21:10 |
| * NobodyCam floods jroll with pm's ..hhehehe | 21:11 | |
| Nisha | but if spec is not approved under Nova, then we have to use workaround for capabilities | 21:11 |
| jroll | what | 21:11 |
| jroll | sigh | 21:12 |
| Nisha | jroll: thats a generic change applicable to all who wants to use list as a value to a capability | 21:12 |
| jroll | we need to do one or the other first | 21:12 |
| jroll | I don't care which | 21:12 |
| Nisha | :) | 21:12 |
| kfox1111 | Is there a place to download some test images rather then having to build them all myself? | 21:12 |
| kfox1111 | like deploy and cirros or something? | 21:13 |
| Nisha | jroll I just wanted to ask because i tend to discover properties along with node-properties discovery (whatever can be done via discovery) | 21:13 |
| Nisha | jroll: so wanted to know the approach which will be acceptable here ;) | 21:14 |
| Nisha | the above stmt is for driver part | 21:14 |
| Nisha | not for generic one | 21:14 |
| jroll | Nisha: right, ok | 21:14 |
| NobodyCam | kfox1111: I'm sure there are. I just don't have a link handy to offer | 21:15 |
| *** david-lyle has joined #openstack-ironic | 21:15 | |
| kfox1111 | k. thanks. | 21:15 |
| *** romcheg has quit IRC | 21:16 | |
| kfox1111 | I guess we will race for it... started the image build. I'll see if Ican find a prebuilt one in less time. :) | 21:17 |
| devananda | Nisha: some comments posted on 133534 | 21:18 |
| NobodyCam | :) kfox1111 if you do find them lets us know ware :) | 21:18 |
| Nisha | devananda: ok. just reading through your comments on discovery spec | 21:18 |
| Nisha | devananda: thanks. | 21:18 |
| Nisha | devananda: one query i had on one comment | 21:19 |
| devananda | jroll, Nisha: +1 to WIP the Nova specs until we have concensus on the approach Ironic wants to take | 21:19 |
| kfox1111 | hah. building it from scratch was quicker. :) impressive. :) | 21:20 |
| devananda | oh, food. I should eat | 21:20 |
| NobodyCam | food, mmmmm, bbiab | 21:20 |
| Nisha | with ilo drivers i tend to discover mandatory properties + hardware capabilities. So is it fine to do so. anyway the ilo driver spec is already dependent on the generic spec. I have proposed the capability part in that | 21:20 |
| kfox1111 | but, fail getting the kernel. modprobe: FATAL: Module nbd not found. :/ | 21:22 |
| NobodyCam | modprob nbd | 21:23 |
| kfox1111 | thats the problem. no module. | 21:23 |
| Nisha | jroll: devananda thanks for the imputs...and reviews | 21:23 |
| NobodyCam | kfox1111: ubuntu? | 21:24 |
| kfox1111 | rhel7 | 21:24 |
| Nisha | will now go for sleep...very-very late here now...will take atleast 2 hrs sleep before going to office :) | 21:24 |
| jroll | no problem Nisha | 21:25 |
| NobodyCam | I think the package is nbd-server ... but am not 100% on that | 21:27 |
| kfox1111 | i installed nbd package. it has server and client in it. but no kernel module. | 21:28 |
| kfox1111 | I'm wondering if they build it into the kernel. | 21:28 |
| NobodyCam | humm maybe libguestfs according to http://linux.web.cern.ch/linux/rhel/rhel7/Red_Hat_Enterprise_Linux-7-7.0_Release_Notes-en-US.pdf | 21:29 |
| kfox1111 | not obvious how I get the script to work with it... | 21:33 |
| kfox1111 | all I need to do is pull the kernel/initrd out right? | 21:33 |
| *** romcheg has joined #openstack-ironic | 21:33 | |
| NobodyCam | kfox1111: ya | 21:38 |
| *** igordcard has joined #openstack-ironic | 21:38 | |
| * NobodyCam heads back to the food... bbiab | 21:39 | |
| kfox1111 | a little sketchy, but this worked: guestfish -a /tmp/my-image.qcow2 run : mount /dev/sda / : download /boot/vmlinuz-3.13.0-39-generic vmlinuz : download /boot/initrd.img-3.13.0-39-generic initrd | 21:39 |
| kfox1111 | thx. | 21:39 |
| *** penick has joined #openstack-ironic | 21:46 | |
| *** eghobo has quit IRC | 21:46 | |
| *** penick has quit IRC | 21:47 | |
| openstackgerrit | Ruby Loo proposed openstack/ironic: Update 'Introduction to Ironic' document https://review.openstack.org/136136 | 21:48 |
| *** eghobo has joined #openstack-ironic | 21:52 | |
| *** Marga_ has joined #openstack-ironic | 21:52 | |
| kfox1111 | I'm a little confused. with ironic node-create, what does the pxe_root_gb do? why isnt it the same as local_gb? | 21:54 |
| *** penick has joined #openstack-ironic | 22:03 | |
| *** jgrimm is now known as zz_jgrimm | 22:06 | |
| *** rwsu has quit IRC | 22:12 | |
| jroll | who wants a super easy review? :) https://review.openstack.org/#/c/135999/ | 22:14 |
| *** rwsu_ has joined #openstack-ironic | 22:15 | |
| *** ryanpetrello has quit IRC | 22:15 | |
| *** mjturek has quit IRC | 22:15 | |
| *** linggao has quit IRC | 22:18 | |
| *** rwsu_ has quit IRC | 22:19 | |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic-specs: Discover node properties using new CLI node-discover-properties https://review.openstack.org/100951 | 22:20 |
| anderbubble | I am a university HPC engineer, and would like to use Ironic to provision our next hpc cluster. What is the best way to get started / prepared for actually contributing back to the documentation or code, as we start using it? | 22:22 |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic-specs: Discover node properties using new CLI node-discover-properties https://review.openstack.org/100951 | 22:22 |
| jroll | anderbubble: start with our docs :) http://docs.openstack.org/developer/ironic/#admin-guide | 22:22 |
| jroll | you'll want to read "introduction to ironic" | 22:23 |
| jroll | and then probably either "deploying ironic with devstack" or "bare metal service installation guide" | 22:23 |
| openstackgerrit | Merged openstack/ironic-python-agent: Support configdrive from HTTP(S) https://review.openstack.org/135792 | 22:23 |
| anderbubble | I’ve already started reading docs, and am working on installing Ironic on my OpenStack test nodes. I think I was looking for https://wiki.openstack.org/wiki/How_To_Contribute :) | 22:26 |
| *** jistr|mtg has quit IRC | 22:27 | |
| jroll | oh, perfect :) | 22:27 |
| jroll | yes, that's what you want for contributing patches | 22:27 |
| jroll | sorry, I misread | 22:27 |
| * NobodyCam is back | 22:28 | |
| NobodyCam | anderbubble: awesome-ness :) | 22:28 |
| anderbubble | I went to an OpenStack meetup this week that touched on Ironic, and there was a bit of trepidation about the usability of Ironic at this point. (Also some in the vids from the summit.) I just wanted to make sure I was prepared to actually help make things better, if it didn’t work for us out-of-the-gate. | 22:29 |
| NobodyCam | jroll: have you looked at a the job that failed? | 22:29 |
| anderbubble | Speaking of which: do I actually *have* to have IPMI to run Ironic, or can I mess around with it on some desktops first? (I’ve got 4-5 real servers to test; but our dc manager hasn’t provisioned them for me yet.) | 22:29 |
| jroll | NobodyCam: gah, tripleo | 22:30 |
| jroll | NobodyCam: 2014-11-20 16:32:32.574 | No distributions at all found for PrettyTable>=0.7,<0.8 (from python-cinderclient) | 22:30 |
| jroll | seems unrelated :P | 22:30 |
| NobodyCam | jroll: lol | 22:30 |
| NobodyCam | anderbubble: we have folks using ironic in production right now | 22:30 |
| anderbubble | NobodyCam, Oh, I know; but it sounded like there’s been at least some dev-level work, at least in the installations I’ve seen. Not saying it doesn’t work: just that it might take some work, too. | 22:31 |
| anderbubble | If I can get OpenStack in here, maybe I can eventually get OCP in here, too. Then my evil plan will be complete! ;) | 22:32 |
| NobodyCam | anderbubble: we have the pxe_ssh driver for testing envs | 22:32 |
| jroll | ocp ftw :D | 22:33 |
| NobodyCam | jroll: robocop ref? | 22:33 |
| jroll | ocp is opencompute | 22:33 |
| NobodyCam | oh | 22:33 |
| NobodyCam | hehehe | 22:33 |
| jroll | :) | 22:33 |
| anderbubble | NobodyCam, what does the pxe_ssh driver do differently from the default (and which is the default?) | 22:34 |
| jroll | anderbubble: it uses the ssh power driver | 22:35 |
| anderbubble | so it runs ipmitool in the tenant environment over ssh? | 22:35 |
| jroll | which uses ssh and cli commands (virsh, VBoxManage, etc) to manage VMs that simulate bare metal machines | 22:35 |
| NobodyCam | anderbubble: our refference driver is the pxe_ipmitool driver, the pxe_ssh driver is use to control Vbox,Vmware, and vish vms over ssh | 22:35 |
| NobodyCam | anderbubble: it uses ssh and cli commands | 22:36 |
| NobodyCam | as jroll said :-p | 22:36 |
| NobodyCam | hehehe | 22:36 |
| anderbubble | Oh, I see. I have a pile of desktops and was wanting to test Ironic on them with *real* baremetal provisioning. They do PXE, but not IPMI, and I didn’t know if I could actually do a (non-production) Ironic deployment on them, or if I’d have to wait for the rackmounts to see it for real. | 22:37 |
| *** Nisha has quit IRC | 22:38 | |
| *** andreykurilin_ has quit IRC | 22:38 | |
| jroll | there's like... a log power driver | 22:38 |
| jroll | which logs a message for the operator to go push a button | 22:38 |
| jroll | I don't think it's in tree but it's somewhere | 22:38 |
| anderbubble | Nice. I’m more concerned with the polling loop that tries to ascertain the *state* of the power. | 22:38 |
| jroll | yeah, the log driver would make that a no-op afaik | 22:39 |
| NobodyCam | jroll: +2 & +a'd | 22:39 |
| jroll | woohoo | 22:39 |
| * jroll is landing all the things this week | 22:39 | |
| jroll | thanks | 22:39 |
| NobodyCam | jroll: we have a meat-space driver too | 22:39 |
| NobodyCam | lol | 22:39 |
| jroll | oh, did it change names? :P | 22:40 |
| NobodyCam | thou I don't know where it is in use | 22:40 |
| anderbubble | need url! :) | 22:40 |
| jroll | anderbubble: if you want to buy some NUCs, there's a lego power driver for those :D | 22:40 |
| NobodyCam | lol | 22:40 |
| jroll | lego mindstorm, that is | 22:40 |
| jroll | https://twitter.com/devananda/status/529333494787227648 | 22:41 |
| jroll | ^ lego driver in action | 22:41 |
| devananda | jroll: there's also an AMT driver proposed to Ironic now | 22:42 |
| jroll | yep | 22:42 |
| PaulCzar | has anyone done a nova driver for ironic ? so that you can boot instances in a different cloud using ironic | 22:42 |
| devananda | anderbubble: if those desktops support AMT / vPro, I would suggest trying out the proposed AMT power driver | 22:42 |
| PaulCzar | useful for testing purposes | 22:43 |
| devananda | PaulCzar: erm? that's something we did, like, a while back. It's been in Nova | 22:43 |
| devananda | PaulCzar: or do you mean a driver within Ironic that talks to the-nova-over-there ? | 22:43 |
| jroll | sounds like the latter | 22:43 |
| PaulCzar | that second thing | 22:43 |
| jroll | that's interesting | 22:43 |
| devananda | oh. nope | 22:43 |
| PaulCzar | instead of virsh or vbox | 22:44 |
| jroll | I feel like someone was talking about that a while back | 22:44 |
| devananda | PaulCzar: look up quitupleO | 22:44 |
| PaulCzar | for the agent_ssh | 22:44 |
| devananda | I've pushed very very strongly against that | 22:44 |
| PaulCzar | I have enough opinions out OOO to not want to suffer through reading about OOOO | 22:44 |
| jroll | where does it end | 22:44 |
| jroll | lol | 22:44 |
| jroll | if you really wnt to cry, PaulCzar: http://specs.openstack.org/openstack/tripleo-specs/specs/juno/tripleo-on-openstack.html | 22:45 |
| PaulCzar | stahp | 22:45 |
| anderbubble | devananda, This is the first I’ve heard of AMT or vPro, so I don’t know. They’re Dell OptiPlexes. Might be that they require a discrete card, but I really have no idea. | 22:45 |
| jroll | I'm just now reading it for the first time | 22:45 |
| PaulCzar | on docker in vagrant on ec2 | 22:45 |
| anderbubble | Looks like the PXEAndSSHDriver would do me good, though. | 22:45 |
| jroll | lifeless: do you know where we can find the meatspace driver? | 22:45 |
| jroll | slash log driver? | 22:45 |
| devananda | anderbubble: the SSH class of drivers will SSH to a host and run virsh/vbox/vmware commands on it -- works for testing with a local hypervisor that emulates baremetal | 22:46 |
| NobodyCam | jroll: there was this awhile back https://review.openstack.org/#/c/98950/ | 22:46 |
| anderbubble | Oh, right. grumble*grumble | 22:46 |
| jroll | NobodyCam: aha, that's it | 22:46 |
| jroll | anderbubble: you should check that thing out | 22:46 |
| * NobodyCam it never landed | 22:47 | |
| devananda | anderbubble: quick googling reveals this: http://www.intel.com/content/www/us/en/architecture-and-technology/intel-active-management-technology.html | 22:47 |
| anderbubble | devanada, I’ll look into it. Thanks. | 22:47 |
| jroll | NobodyCam: wait, lol, that doesn't even log | 22:47 |
| devananda | if the desktops you're tinkering with support that, then check out https://review.openstack.org/135184 | 22:47 |
| jroll | isn't that the whole point? | 22:47 |
| anderbubble | devanada, will do. Thanks. | 22:48 |
| NobodyCam | jroll: humm maybe thats why it never landed lol | 22:50 |
| jroll | hahaha | 22:50 |
| jroll | brb | 22:50 |
| devananda | adam_g: https://review.openstack.org/#/c/130146/3 -- thoughts? | 22:50 |
| devananda | adam_g: looking at the history there, I think we should (have) approve(d) it (back in october) | 22:52 |
| *** spandhe has quit IRC | 22:52 | |
| adam_g | devananda, digging for context, sec | 22:53 |
| adam_g | devananda, yeah--i think if that branch were active, people would have had trouble merging things back then. i dont see any harm in merging it now to keep synced | 22:58 |
| NobodyCam | devananda: do you know if testtools 1.4.0 is ok? I asked a while back but never followed up on getting the answer | 22:58 |
| NobodyCam | fyi: https://review.openstack.org/#/c/135985/1/test-requirements.txt | 22:59 |
| NobodyCam | nm | 22:59 |
| *** spandhe has joined #openstack-ironic | 23:00 | |
| *** eghobo has quit IRC | 23:00 | |
| devananda | adam_g: +A'd | 23:02 |
| *** igordcard has quit IRC | 23:02 | |
| adam_g | so i'm playing with this multinode devstack stuff that people are working on, trying to figure out how we might be able to expand our devstack infra to include a second conductor and more VM capacity | 23:04 |
| devananda | hmm | 23:04 |
| adam_g | we could potentially test hash ring stuff at some point | 23:04 |
| devananda | neat! | 23:04 |
| devananda | I didn't realize the multinode-devstack stuff would be possible in infra | 23:04 |
| devananda | *in the gate | 23:04 |
| adam_g | devananda, me either, but now it is :) | 23:05 |
| adam_g | clarkb and jogo have it going with a nova-network multi_host setup via some gre overlay networks, wondering how we might adapt it to fit the ironic setup | 23:06 |
| *** athomas has quit IRC | 23:18 | |
| kfox1111 | hmm..... I think I'm close here. nova's failing to schedule due to "ComputeCapabilitiesFilter returned 0 hosts" | 23:19 |
| kfox1111 | any idea? | 23:19 |
| NobodyCam | kfox1111: node and flavor spec's need to match | 23:19 |
| kfox1111 | node properties {u'memory_mb': u'65536', u'cpu_arch': u'x86_64', u'local_gb': u'1000', | | 23:20 |
| kfox1111 | | | u'cpus': u'32'} | 23:20 |
| kfox1111 | | ccf808c4-b5ed-4ebb-9076-6285679d8eeb | my-baremetal-flavor | 62000 | 900 | 0 | | 32 | 1.0 | True | | 23:20 |
| jroll | are you using ExactRamFilter? | 23:20 |
| kfox1111 | hmm... | 23:20 |
| jroll | because those don't match RAM size | 23:20 |
| jroll | same for ExactDiskFilter | 23:20 |
| NobodyCam | kfox1111: see http://docs.openstack.org/developer/ironic/deploy/install-guide.html#flavor-creation | 23:21 |
| kfox1111 | in the baremetal_scheduler_default_filters? | 23:21 |
| devananda | sirushti: ping | 23:22 |
| kfox1111 | I just changed memory back to 65536, and filter RamFilter returned 0 hosts | 23:23 |
| kfox1111 | hmm.. let me double check the nova config vs the doc. | 23:24 |
| NobodyCam | :) | 23:24 |
| NobodyCam | bbiafm.. need fresh air (smoke) | 23:25 |
| devananda | kfox1111: check output of "nova hypervisor-list" and "nova hypervisor-show ###" | 23:25 |
| devananda | kfox1111: make sure the stats reported match what Ironic shows for the node.properties, and both should match the flavor | 23:25 |
| * devananda thinks we should add a section to the docs showing how to troubleshoot this, and explaining what "nova hypervisor-show" means with Ironic | 23:26 | |
| kfox1111 | the configs match. | 23:26 |
| *** romcheg has quit IRC | 23:26 | |
| *** romcheg has joined #openstack-ironic | 23:27 | |
| *** alexpilotti has quit IRC | 23:27 | |
| *** romcheg has quit IRC | 23:27 | |
| kfox1111 | cores 32, free_disk_gb 877, free_ram_mb 63736.... is that part of it? memory_mb 64248 memory_mb_used 512 | 23:27 |
| devananda | used 512 -- that's odd | 23:27 |
| kfox1111 | vcpus 32 vcpus_used 0 | 23:28 |
| kfox1111 | hypervisor_type QEMU? | 23:28 |
| devananda | so a flavor requiring 900 gb of disk is not going to match a host with 877 | 23:29 |
| devananda | also, why is it showing ANY memory_mb_used at all? | 23:29 |
| devananda | I suspect a misconfiguration somewhere, but need more info | 23:29 |
| kfox1111 | yeah | 23:29 |
| *** anderbubble has quit IRC | 23:30 | |
| devananda | when no instance is deployed to the node, it should report no resources used at all | 23:30 |
| devananda | cnoversely, when an instance IS deployed, it should report ALL resources used | 23:30 |
| kfox1111 | yeah. and where did that qemu thing come from? | 23:32 |
| devananda | what qemu thing? | 23:33 |
| devananda | kfox1111: hm. someone was in here yesterday asking the same thing. turned out to be an auth problem -- nova couldn't connect to the Ironic service with the credentials it was given | 23:34 |
| kfox1111 | hmm... I'll take a look at that. | 23:34 |
| kfox1111 | here's my nova conf by the way: http://pastebin.com/q6kh7smP | 23:35 |
| kfox1111 | good call. there is a missing admin_tenant_name in [ironic] | 23:36 |
| devananda | yup | 23:36 |
| devananda | look at the nova-compute log file -- it should indicate the failure | 23:37 |
| * NobodyCam is back | 23:37 | |
| kfox1111 | yup. still failed... | 23:38 |
| devananda | kfox1111: what's the log say? | 23:39 |
| kfox1111 | its failing authentication when I try and boot an instance. its fine before that. | 23:40 |
| kfox1111 | [-] Error during ClusteredComputeManager.update_available_resource: The request you have made requires authentication. (HTTP 401) | 23:40 |
| kfox1111 | let me double check that config... | 23:40 |
| devananda | vdrok: how will non-glance image support interact with things like https://review.openstack.org/#/c/97150/15/specs/kilo/whole-disk-image.rst wherein we need a property from the glance image? | 23:40 |
| *** Marga_ has quit IRC | 23:40 | |
| NobodyCam | oh good question devananda :) | 23:41 |
| devananda | kfox1111: try using ironic_host_manager | 23:41 |
| kfox1111 | there we go. tenant was stil wrong. now its over to ironic. | 23:41 |
| devananda | nm then :) | 23:42 |
| kfox1111 | PortNotFree: No free port available for instance. | 23:42 |
| NobodyCam | kfox1111: did you add a port to your ironic node? | 23:42 |
| kfox1111 | didn't see instructions telling me to do that. :) | 23:42 |
| devananda | ironic list-ports <NODE UUID> | 23:42 |
| devananda | you need to inform Ironic of the MAC address(es) of the servers it's managing | 23:42 |
| devananda | so that it can pass the right info to neutron to set up the networking | 23:43 |
| kfox1111 | ah. that makes sense. | 23:43 |
| * devananda wonders why that's not in our docs | 23:43 | |
| kfox1111 | hmm.. yeah. its not in the node-show. ok. lets see... | 23:43 |
| *** ryanpetrello has joined #openstack-ironic | 23:45 | |
| kfox1111 | nice. ok. next up tftp config bits.... just a min. | 23:46 |
| NobodyCam | kfox1111: about half way down this page http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html See: Here is an example walkthrough of creating a node section | 23:46 |
| kfox1111 | ok. thanks. | 23:48 |
| *** igordcard has joined #openstack-ironic | 23:49 | |
| *** spandhe has quit IRC | 23:49 | |
| *** ryanpetrello has quit IRC | 23:52 | |
| kfox1111 | sed -i 's/\(disable[ \t]*=[ \t]\).*/\1no/' /etc/xinetd.d/tftp | 23:52 |
| kfox1111 | :/ | 23:52 |
| *** Haomeng|2 has joined #openstack-ironic | 23:55 | |
| NobodyCam | enabling tftp thru xinetd | 23:55 |
| *** ryanpetrello has joined #openstack-ironic | 23:55 | |
| *** Haomeng has quit IRC | 23:56 | |
| kfox1111 | its listening... but not responding... iptables is ok.... | 23:56 |
| NobodyCam | anything in syslog do you see the requests comming in? | 23:57 |
| kfox1111 | hmm... no. | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!