| *** mdurnosvistov_ has quit IRC | 00:34 | |
| *** nosnos has joined #openstack-ironic | 01:37 | |
| *** nosnos has quit IRC | 01:38 | |
| *** nosnos has joined #openstack-ironic | 01:40 | |
| *** jbjohnso has quit IRC | 01:58 | |
| *** lazy_prince has quit IRC | 02:14 | |
| *** killer_prince has joined #openstack-ironic | 02:15 | |
| *** murzer has quit IRC | 02:17 | |
| *** lazy_prince has joined #openstack-ironic | 03:23 | |
| *** lazy_prince has quit IRC | 03:28 | |
| *** killer_prince is now known as lazy_prince | 03:31 | |
| *** lazy_prince2 has joined #openstack-ironic | 04:45 | |
| *** lazy_prince2 is now known as killer_prince | 04:45 | |
| *** rameshg87 has joined #openstack-ironic | 04:56 | |
| *** saju_m has joined #openstack-ironic | 04:59 | |
| *** nosnos has quit IRC | 05:39 | |
| *** nosnos has joined #openstack-ironic | 05:39 | |
| *** saju_m has quit IRC | 05:54 | |
| *** saju_m has joined #openstack-ironic | 05:55 | |
| *** loki_ has joined #openstack-ironic | 05:56 | |
| *** saju_m has quit IRC | 06:00 | |
| *** saju_m has joined #openstack-ironic | 06:02 | |
| openstackgerrit | Jenkins proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/78862 | 06:07 |
|---|---|---|
| *** nosnos_ has joined #openstack-ironic | 06:51 | |
| *** nosnos has quit IRC | 06:51 | |
| *** nosnos has joined #openstack-ironic | 06:56 | |
| *** nosnos_ has quit IRC | 06:58 | |
| *** nosnos has quit IRC | 07:05 | |
| *** nosnos has joined #openstack-ironic | 07:08 | |
| *** romcheg has joined #openstack-ironic | 07:11 | |
| *** nosnos has quit IRC | 07:12 | |
| *** nosnos has joined #openstack-ironic | 07:12 | |
| *** max_lobur has joined #openstack-ironic | 07:16 | |
| *** matsuhashi has joined #openstack-ironic | 07:39 | |
| *** max_lobur has quit IRC | 07:39 | |
| *** mdurnosvistov_ has joined #openstack-ironic | 08:12 | |
| *** saju_m has quit IRC | 08:22 | |
| GheRivero | morning ironic | 08:22 |
| *** saju_m has joined #openstack-ironic | 08:28 | |
| romcheg | Morning GheRivero | 08:28 |
| *** ifarkas has joined #openstack-ironic | 08:32 | |
| openstackgerrit | yangxurong proposed a change to openstack/ironic: Sort requirement files in alphabetical order https://review.openstack.org/79276 | 08:40 |
| *** Manishanker has left #openstack-ironic | 08:41 | |
| *** matsuhashi has quit IRC | 08:50 | |
| *** saju_m has quit IRC | 08:50 | |
| *** matsuhashi has joined #openstack-ironic | 08:54 | |
| *** saju_m has joined #openstack-ironic | 09:04 | |
| *** romcheg has left #openstack-ironic | 09:07 | |
| *** saju_m has quit IRC | 09:10 | |
| *** saju_m has joined #openstack-ironic | 09:11 | |
| *** martyntaylor has joined #openstack-ironic | 09:12 | |
| *** derekh has joined #openstack-ironic | 09:12 | |
| *** ndipanoff has joined #openstack-ironic | 09:15 | |
| *** jistr has joined #openstack-ironic | 09:15 | |
| *** lucasagomes has joined #openstack-ironic | 09:16 | |
| *** athomas has joined #openstack-ironic | 09:31 | |
| *** ndipanoff is now known as ndipanov | 09:49 | |
| *** martyntaylor has quit IRC | 10:07 | |
| *** jrist has quit IRC | 10:08 | |
| *** jrist has joined #openstack-ironic | 10:21 | |
| *** martyntaylor has joined #openstack-ironic | 10:25 | |
| *** romcheg has joined #openstack-ironic | 10:36 | |
| GheRivero | lucasagomes: About the multiplex driver, I remember that you suggested to use decorators for something, but I don't remember for what. | 10:45 |
| lucasagomes | GheRivero, oh it was an suggestion but we dropped it | 10:45 |
| GheRivero | ok. thanks | 10:46 |
| lucasagomes | I suggested to use some decorators on the methods that will be exposed to the vendorpassthru | 10:46 |
| lucasagomes | but we voted to have something more explicit instead | 10:46 |
| lucasagomes | GheRivero, like the dict you put on ur patch :) | 10:46 |
| GheRivero | oh yes! I remember now. | 10:46 |
| GheRivero | I though I was missing on the patch | 10:47 |
| *** romcheg1 has joined #openstack-ironic | 10:47 | |
| lucasagomes | :) | 10:47 |
| lucasagomes | u grand | 10:47 |
| *** romcheg has quit IRC | 10:47 | |
| *** mdurnosvistov__ has joined #openstack-ironic | 10:48 | |
| *** mdurnosvistov_ has quit IRC | 10:50 | |
| *** romcheg1 has quit IRC | 10:56 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Generic MixinVendorInterface using static mapping https://review.openstack.org/78384 | 10:58 |
| *** saju_m has quit IRC | 11:05 | |
| *** romcheg has joined #openstack-ironic | 11:31 | |
| *** max_lobur has joined #openstack-ironic | 11:38 | |
| *** loki_ has quit IRC | 11:50 | |
| *** killer_prince has quit IRC | 11:56 | |
| *** matsuhashi has quit IRC | 12:11 | |
| *** matsuhashi has joined #openstack-ironic | 12:13 | |
| *** linggao has joined #openstack-ironic | 12:17 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Generic MixinVendorInterface using static mapping https://review.openstack.org/78384 | 12:19 |
| *** lazy_prince is now known as killer_prince | 12:27 | |
| *** mdurnosvistov__ has quit IRC | 12:33 | |
| *** jdob has joined #openstack-ironic | 12:35 | |
| *** lucasagomes is now known as lucas-hungry | 12:39 | |
| *** killer_prince has quit IRC | 12:44 | |
| *** rameshg87 has left #openstack-ironic | 12:48 | |
| *** killer_prince has joined #openstack-ironic | 12:51 | |
| *** romcheg1 has joined #openstack-ironic | 13:05 | |
| *** romcheg has quit IRC | 13:05 | |
| *** matty_dubs|gone is now known as matty_dubs | 13:12 | |
| matty_dubs | Good morning, Ironic! | 13:14 |
| *** rustlebee is now known as russellb | 13:26 | |
| *** matsuhashi has quit IRC | 13:32 | |
| *** ndipanov_ has joined #openstack-ironic | 13:33 | |
| *** romcheg1 has left #openstack-ironic | 13:33 | |
| *** ndipanov_ has quit IRC | 13:33 | |
| *** ndipanov_ has joined #openstack-ironic | 13:34 | |
| *** ndipanov has quit IRC | 13:36 | |
| *** nosnos has quit IRC | 13:38 | |
| *** lucas-hungry is now known as lucasagomes | 13:39 | |
| *** killer_prince is now known as lazy_prince | 13:46 | |
| *** jgrimm has quit IRC | 13:51 | |
| *** ndipanov_ is now known as ndipanov | 13:53 | |
| *** ndipanov has quit IRC | 13:53 | |
| *** ndipanov has joined #openstack-ironic | 13:54 | |
| *** jbjohnso has joined #openstack-ironic | 14:13 | |
| lucasagomes | lifeless, ping re disk partitioning in Ironic | 14:16 |
| lucasagomes | lifeless, was thinking, right now we have 2 layouts: | 14:16 |
| lucasagomes | if ephemeral is not present, root will be the first partition and then swap | 14:17 |
| lucasagomes | if ephemeral is present, ephemeral will be the first, then swap, then root | 14:17 |
| lucasagomes | lifeless, for consistency sake, I think make sense to put root _always_ as the last partition, any objections? | 14:18 |
| *** rwsu has joined #openstack-ironic | 14:36 | |
| *** davidlenwell has quit IRC | 14:48 | |
| *** martyntaylor has quit IRC | 14:49 | |
| BlartVersenwald | Good morning Ironic | 14:49 |
| *** BlartVersenwald is now known as NobodyCam | 14:49 | |
| openstackgerrit | Jenkins proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/79334 | 14:49 |
| *** davidlenwell_ has joined #openstack-ironic | 14:49 | |
| *** martyntaylor has joined #openstack-ironic | 14:51 | |
| *** blamar has joined #openstack-ironic | 15:00 | |
| *** ndipanov has quit IRC | 15:02 | |
| *** blamar has quit IRC | 15:02 | |
| *** tatyana has joined #openstack-ironic | 15:02 | |
| *** blamar has joined #openstack-ironic | 15:02 | |
| *** jgrimm has joined #openstack-ironic | 15:03 | |
| *** ndipanov has joined #openstack-ironic | 15:06 | |
| *** rloo has joined #openstack-ironic | 15:11 | |
| *** datajerk1 has joined #openstack-ironic | 15:30 | |
| *** datajerk has quit IRC | 15:30 | |
| NobodyCam | slow morning today! | 15:52 |
| lucasagomes | morning NobodyCam matty_dubs | 15:53 |
| NobodyCam | morning lucasagomes | 15:53 |
| lucasagomes | :) | 15:54 |
| lucasagomes | NobodyCam, on the road already? | 15:54 |
| *** lazy_prince is now known as killer_prince | 15:54 | |
| NobodyCam | lucasagomes: home already | 15:55 |
| NobodyCam | :) | 15:55 |
| lucasagomes | :D | 15:55 |
| lucasagomes | awesome | 15:55 |
| NobodyCam | :) | 15:55 |
| *** Hefeweizen has quit IRC | 16:00 | |
| matty_dubs | Howdy lucasagomes! | 16:02 |
| NobodyCam | Morning matty_dubs :) | 16:03 |
| lucasagomes | :) | 16:04 |
| *** hemna__ has quit IRC | 16:08 | |
| *** ifarkas has quit IRC | 16:08 | |
| *** vkozhukalov has joined #openstack-ironic | 16:08 | |
| openstackgerrit | A change was merged to openstack/ironic: Move v1 API tests to separate v1 directory https://review.openstack.org/78918 | 16:09 |
| *** Hefeweizen has joined #openstack-ironic | 16:18 | |
| *** matty_dubs is now known as matty_dubs|lunch | 16:20 | |
| NobodyCam | bbt ...brb | 16:21 |
| *** jgrimm has quit IRC | 16:27 | |
| *** harlowja has joined #openstack-ironic | 16:49 | |
| * NobodyCam shutters with every he uses --trash-my-machine | 16:52 | |
| NobodyCam | :-p | 16:52 |
| * NobodyCam shutters with every use of --trash-my-machine | 16:52 | |
| jroll | what's the HA model for the ironic API? spin up X servers and throw them behind a load balancer should work, right? | 16:53 |
| NobodyCam | jroll: yep that should work | 16:54 |
| jroll | cool, thanks | 16:54 |
| *** killer_prince is now known as lazy_prince | 16:57 | |
| *** ifarkas has joined #openstack-ironic | 17:02 | |
| *** matty_dubs|lunch is now known as matty_dubs | 17:07 | |
| *** jistr has quit IRC | 17:16 | |
| *** eghobo has joined #openstack-ironic | 17:23 | |
| *** dhellmann is now known as dhellmann_ | 17:23 | |
| NobodyCam | lucasagomes: quick question you may know the answer to. icli.node.update returns a node object, correct? | 17:24 |
| lucasagomes | NobodyCam, yes | 17:25 |
| lucasagomes | it returns the updated node | 17:25 |
| NobodyCam | ok :) | 17:25 |
| NobodyCam | going to see if I can do a little clenup while fighting with devtest | 17:26 |
| NobodyCam | :) | 17:26 |
| *** martyntaylor has quit IRC | 17:26 | |
| NobodyCam | oh are we meeting today, are people still travling? | 17:26 |
| lucasagomes | NobodyCam, idk, is devananda here? | 17:29 |
| NobodyCam | he's running a tad late this morning, he'll be on around 11 | 17:29 |
| NobodyCam | (pst ) | 17:29 |
| lucasagomes | NobodyCam, btw I have a small fix for the driver, the retry_on_conflict doesn't work anymore | 17:29 |
| lucasagomes | because it nows returns httpserviceunavailable | 17:29 |
| lucasagomes | for temporary failures | 17:30 |
| lucasagomes | I mean temporary locks etc | 17:30 |
| lucasagomes | not failures | 17:30 |
| lucasagomes | NobodyCam, ic | 17:30 |
| NobodyCam | do you want to push that up or past me the change? | 17:30 |
| lucasagomes | NobodyCam, will do, testing it right now | 17:30 |
| NobodyCam | :) | 17:30 |
| lucasagomes | just in case | 17:31 |
| lucasagomes | saw my deploy failing because of that | 17:31 |
| lucasagomes | quicked fixed lemme see | 17:31 |
| NobodyCam | ahh :) I'm tring to get my env working again | 17:31 |
| NobodyCam | :-p | 17:31 |
| *** rloo has quit IRC | 17:31 | |
| lucasagomes | NobodyCam, tripleo or devstack? | 17:31 |
| NobodyCam | ooo | 17:32 |
| *** rloo has joined #openstack-ironic | 17:32 | |
| lucasagomes | ack | 17:32 |
| lucasagomes | NobodyCam, btw, I put up two bp today | 17:32 |
| lucasagomes | if u wanna take a look | 17:32 |
| * lucasagomes go grab the links | 17:32 | |
| lucasagomes | NobodyCam, https://blueprints.launchpad.net/ironic/+spec/advanced-partition-manager | 17:32 |
| lucasagomes | NobodyCam, https://blueprints.launchpad.net/ironic/+spec/ipxe-boot | 17:32 |
| NobodyCam | :) #1 makes good sense to me :) I havn't use pyparted but seems like the way to go | 17:34 |
| lucasagomes | NobodyCam, :) or udisks | 17:34 |
| NobodyCam | udisks would have two dep's udsik and python-udisks? | 17:35 |
| NobodyCam | or just python-udisks | 17:35 |
| NobodyCam | (/me has not looked) | 17:35 |
| lucasagomes | NobodyCam, python-udisks is just a wrapper | 17:35 |
| lucasagomes | udisks is a daemon that does manage disks | 17:36 |
| lucasagomes | and it has a dbus interface | 17:36 |
| lucasagomes | the python-udisks is a wrapper to manage udisks like objects | 17:36 |
| lucasagomes | but all it does is do dbus calls to the udisks daemon | 17:36 |
| NobodyCam | ya, parted is installed by default on most distros no? | 17:36 |
| lucasagomes | yup | 17:36 |
| NobodyCam | :) | 17:36 |
| lucasagomes | udisks is instealled by default on fedora as well | 17:37 |
| lucasagomes | not sure about ubuntu | 17:37 |
| NobodyCam | dbus interface is neat sounding | 17:37 |
| lucasagomes | yeah | 17:37 |
| lucasagomes | I like it | 17:37 |
| NobodyCam | ahh and I have it on my ubuntu | 17:38 |
| NobodyCam | so | 17:38 |
| NobodyCam | :) | 17:38 |
| lucasagomes | :D | 17:38 |
| lucasagomes | python-udisks is not totally complete tho, I gotta add more stuff to it | 17:38 |
| NobodyCam | Command python setup.py egg_info failed with error code 1 in /opt/stack/venvs/python-glanceclient/build/cryptography | 17:38 |
| lucasagomes | but the basic interfaces are there I believe | 17:38 |
| *** ndipanov is now known as ndipanov_gone | 17:42 | |
| *** tatyana has quit IRC | 17:42 | |
| *** martyntaylor has joined #openstack-ironic | 17:43 | |
| NobodyCam | quick walkies... | 17:47 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Nova ironic driver to retry on HTTP 503 https://review.openstack.org/79381 | 17:48 |
| lucasagomes | NobodyCam, ^ | 17:48 |
| lucasagomes | enjoy walkies | 17:48 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Nova ironic driver to retry on HTTP 503 https://review.openstack.org/79381 | 17:50 |
| *** linggao has quit IRC | 17:51 | |
| *** derekh has quit IRC | 17:51 | |
| *** linggao has joined #openstack-ironic | 17:54 | |
| NobodyCam | :) | 17:57 |
| *** athomas has quit IRC | 17:59 | |
| *** dwalleck has joined #openstack-ironic | 17:59 | |
| NobodyCam | max_lobur: you around? | 17:59 |
| jroll | so, meeting? | 18:01 |
| NobodyCam | jroll: daylight savings time | 18:01 |
| NobodyCam | meeting at noon | 18:01 |
| jroll | ohhhh right | 18:02 |
| NobodyCam | or 1900 gmt | 18:02 |
| NobodyCam | :) | 18:02 |
| jroll | thanks :) | 18:02 |
| NobodyCam | always gets me too | 18:02 |
| lifeless | lucasagomes: so we did that thing in nova | 18:04 |
| lucasagomes | lifeless, ack, yeah it looks more consistent | 18:04 |
| lifeless | lucasagomes: to allow manual partitioning of the remainder of the disk without cloud-init taking over, when no ephemeral was used. | 18:04 |
| lucasagomes | lifeless, a-ha great | 18:05 |
| lifeless | lucasagomes: I mean, we did the two layouts in nova-bm for ^ reason | 18:05 |
| lucasagomes | lifeless, got it makes sense | 18:05 |
| lucasagomes | lifeless, also when u get a time I would appreciate ur input at https://blueprints.launchpad.net/ironic/+spec/advanced-partition-manager | 18:05 |
| lifeless | lucasagomes: however, that was before we had ephemeral support sorted, and the reason we kept the old layuot was to make dealing with the transition easy | 18:05 |
| lifeless | lucasagomes: so, +1 on making it more consistent now | 18:05 |
| lucasagomes | lifeless, ack I will work on a patch for that | 18:06 |
| * devananda reads scrollback | 18:06 | |
| lifeless | lucasagomes: I don't think we need to keep compat in this regard with nova baremetal - but worth thinking about the implications for upgrades from nova-bm | 18:06 |
| devananda | morning, all | 18:06 |
| NobodyCam | morning devananda :) | 18:06 |
| lucasagomes | morning devananda | 18:06 |
| lucasagomes | lifeless, yeah, it's not a critical thing | 18:06 |
| lucasagomes | but as I'm also working on get rid of the swap partition | 18:06 |
| dwalleck | Ahh, meeting isn't for another hour? Silly time changes | 18:06 |
| lucasagomes | lifeless, having the partition layout more consistent will make the code cleaner for that as well | 18:07 |
| NobodyCam | devananda: while reading scroll back you wanta glance over https://review.openstack.org/79381 | 18:07 |
| lifeless | lucasagomes: I can't see a spec at that url, just a short paragraph or two | 18:07 |
| NobodyCam | dwalleck: :) yep | 18:07 |
| lucasagomes | lifeless, ouch weird lemme check | 18:07 |
| lifeless | lucasagomes: by upgrade implications, i mean - if you have an instance deployed with nova bm, you upgrade to use ironic, and then you do 'nova rebuild --preserve-ephemeral' | 18:08 |
| lifeless | lucasagomes: we wouldn't want that to be confused about where the root partition is | 18:08 |
| lifeless | lucasagomes: FWIW gpt support is very straight forward in the current structure, and we don't care about reading back which partition is bootable | 18:09 |
| lucasagomes | lifeless, right, that was just an example | 18:09 |
| devananda | lucasagomes: looks to me like _retry_if_service_is_unavailable should be a decorator | 18:09 |
| lifeless | lucasagomes: so the problem statement doesn't make sense to me. | 18:09 |
| devananda | lucasagomes: but the 409 -> 503 change seems correct | 18:10 |
| lifeless | lucasagomes: the idea of bringing udisks into the picture is rather terrifying, it has such a huge dependency tre | 18:10 |
| lifeless | tree | 18:10 |
| devananda | lifeless: I do not want to support a nova-bm -> ironic upgrade path with --preserve-ephemeral | 18:11 |
| lucasagomes | lifeless, right, what I mean in that blueprint is to have a more advanced way to manage the partitions, for e.g I think it was the mirantes folks that were talking about lvm in the tripleo mid-cycle | 18:11 |
| devananda | lifeless: given the testing requirements / overhead for such a thing, when it's not necessary for graduation | 18:11 |
| lucasagomes | so the advanced parition mechanism would just be an "pluggable" way to manage diferent partitioning schemas | 18:12 |
| lucasagomes | devananda, right, I can work on a decorator for that | 18:12 |
| lucasagomes | just fixed as-is cause it was failing in my env | 18:12 |
| devananda | lucasagomes: ack, that's fine | 18:13 |
| lifeless | lucasagomes: so, if folk need it we should support it, but the spec should cover the actual use cases, and proposed implementation, including things like how users specify what they want etc. | 18:13 |
| lifeless | lucasagomes: I'm expecting an etherpad at the 'spec url' on the blueprint page, with lots of detail | 18:13 |
| lucasagomes | lifeless, right, yeah I just put it up today | 18:13 |
| lucasagomes | lifeless, I will work on a more detailed implementation/overview | 18:13 |
| lifeless | lucasagomes: I worry a great deal about the idea of LVM being an Ironic concern | 18:14 |
| lucasagomes | lifeless, well disk managing in general | 18:14 |
| lifeless | lucasagomes: seems like a massive layering violation to me | 18:14 |
| lucasagomes | lifeless, the idea was to have cinder to design the partition layout but ironic would still do the actual partitioning | 18:14 |
| lucasagomes | afaiui | 18:14 |
| lifeless | lucasagomes: perhaps we can discuss in atlanta? | 18:15 |
| lifeless | lucasagomes: seems like a not-yet thing, given our other more pressing needs ;) | 18:15 |
| lucasagomes | lifeless, definitely | 18:15 |
| lucasagomes | lifeless, yup, yeah it's just good to have it up there so people can look at it and put their thoughts on that bp | 18:15 |
| lucasagomes | in atlanta we can propouse a section to talk about it more closely | 18:16 |
| lucasagomes | and try to come up with some solution, either on ironic or cinder | 18:16 |
| lifeless | devananda: re upgrades - unless the current round of vendor products manage to ship with Ironic, TripleO based things will need that upgrade path to work. | 18:16 |
| lifeless | devananda: I'm not saying it should distract from meeting the 'deprecate Nova-BM' focus, but forcing an entire cluster to be redeployed to upgrade seems like its not an upgrade path. | 18:17 |
| lifeless | in that its no easier than just wiping everything and starting over | 18:17 |
| devananda | lifeless: this is one reason why I've been pushing for nova-bm to be frozen | 18:19 |
| lifeless | devananda: The feature is in both nova-bm and ironic | 18:19 |
| devananda | indeed | 18:20 |
| devananda | and ironic can't move that functionality now | 18:20 |
| devananda | s/move/change/ | 18:20 |
| * Shrews wonders why the cherry-pick of 70348 was removed from https://etherpad.openstack.org/p/IronicAndDevstackAgain | 18:20 | |
| lifeless | devananda: so the frozen comment seems a little bit of a non-sequitur | 18:20 |
| lifeless | devananda: it can, change it, it just needs to do so in the same way it would if it was in Ironic previously - gracefully. | 18:20 |
| *** dhellmann_ is now known as dhellmann | 18:23 | |
| NobodyCam | lifeless: quick question for you, have you seen a issue when running devtest on a x64 bast system and building i386 images, where it errors with "cffi.ffiplatform.VerificationError: LinkError: command 'i686-linux-gnu-gcc' failed with exit status 1" I have /usr/bin/x86_64-linux-gnu-gcc but not i686 | 18:23 |
| NobodyCam | s/bast/based/ | 18:24 |
| max_lobur | NobodyCam: hi! around | 18:24 |
| max_lobur | hi Everyone | 18:24 |
| lifeless | NobodyCam: I haven't no. i686-linux-gnu-gcc should be installed in the chroot, not in your host os. | 18:25 |
| lifeless | NobodyCam: use break= or perhaps add a call to 'bash' around where it fails | 18:25 |
| lifeless | NobodyCam: if the logs don't give you enough info | 18:25 |
| NobodyCam | morning max_lobur :) I was just seeing if you around | 18:25 |
| NobodyCam | lifeless: ack :) looking into it now... just saw you in channel figured I'd ask | 18:26 |
| max_lobur | :) | 18:26 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add Node.instance_info field https://review.openstack.org/79396 | 18:26 |
| *** davidlenwell_ is now known as davidlenwell | 18:26 | |
| *** max_lobur has left #openstack-ironic | 18:27 | |
| *** max_lobur has joined #openstack-ironic | 18:27 | |
| *** lucasagomes is now known as lucas-afk | 18:30 | |
| lifeless | the failed to get metadata thing is worrying | 18:32 |
| NobodyCam | ahh hidden in the log: /opt/stack/venvs/python-glanceclient/build/cryptography/cryptography/hazmat/bindings/__pycache__/cryptography/hazmat/bindings/__pycache__/_cffi__x5eaa210axf0ae7e21.o: file not recognized: File truncated | 18:37 |
| devananda | adam_g: still using that hpcs instance for devstack? | 18:39 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add Node.instance_info field https://review.openstack.org/79396 | 18:44 |
| * NobodyCam makes a bagel b4 meeting | 18:46 | |
| openstackgerrit | Ruby Loo proposed a change to openstack/ironic: Implement API to get driver properties https://review.openstack.org/73005 | 18:48 |
| * devananda updates agenda | 18:54 | |
| *** romcheg1 has joined #openstack-ironic | 18:54 | |
| *** lucas-afk is now known as lucasagomes | 18:56 | |
| *** dwalleck_ has joined #openstack-ironic | 18:57 | |
| *** dwalleck has quit IRC | 18:57 | |
| * NobodyCam reviews new agenda and eats his bagel | 18:58 | |
| *** dwalleck has joined #openstack-ironic | 18:59 | |
| * NobodyCam heads over to meeting room | 18:59 | |
| *** dwalleck_ has quit IRC | 19:01 | |
| russell_h | is anyone else seeing an issue with database.connection not being defaulted to anything? | 19:10 |
| russell_h | I don't know what it defaulted to before | 19:10 |
| russell_h | or if this is expected | 19:10 |
| russell_h | but now its None which breaks pretty hard | 19:10 |
| openstackgerrit | A change was merged to openstack/ironic: Nova ironic driver to retry on HTTP 503 https://review.openstack.org/79381 | 19:20 |
| adam_g | devananda: that instance is okay to go away now,t hanks | 19:21 |
| devananda | adam_g: hi! meeting in #openstack-meeting, btw | 19:21 |
| adam_g | devananda: yea! missed the start, but.. | 19:23 |
| openstackgerrit | Ruby Loo proposed a change to openstack/ironic: Implement API to get driver properties https://review.openstack.org/73005 | 19:32 |
| *** vkozhukalov has quit IRC | 19:44 | |
| *** rloo has quit IRC | 19:50 | |
| *** rloo has joined #openstack-ironic | 19:50 | |
| Shrews | adam_g, ifarkas: i think i have a fix for the scheduler issue... testing now | 19:53 |
| *** dwalleck_ has joined #openstack-ironic | 19:56 | |
| *** hmelevsk has joined #openstack-ironic | 19:58 | |
| adam_g | Shrews: cool | 19:58 |
| *** dwalleck has quit IRC | 20:00 | |
| NobodyCam | great meeting all Thank you | 20:00 |
| lucasagomes | devananda, NobodyCam would ask for an action but time's up | 20:01 |
| devananda | hehe | 20:01 |
| lucasagomes | I will send something to the list asking people to give ideas of what methods would be common across vendor drivers | 20:01 |
| NobodyCam | #action lucas to do what he said | 20:01 |
| lucasagomes | and give the set_boot_device as an example | 20:01 |
| lucasagomes | NobodyCam, :P heh | 20:02 |
| lucasagomes | cheers hah | 20:02 |
| NobodyCam | have a good night lucasagomes | 20:02 |
| ifarkas | Shrews, that's great | 20:02 |
| adam_g | Shrews: i was hitting one other issue last week that lib/ironic was doing a restart of nova using stop_nova + start_nova. stop_nova does not stop n-obj, but start_nova tries to start it.. so objectstore fails to start up again (port 3333 already in use) | 20:02 |
| NobodyCam | brb | 20:03 |
| * devananda steps away for a few minutse | 20:04 | |
| *** dwalleck_ has quit IRC | 20:05 | |
| *** dwalleck_ has joined #openstack-ironic | 20:05 | |
| *** dwalleck has joined #openstack-ironic | 20:06 | |
| *** dwalleck_ has quit IRC | 20:10 | |
| lucasagomes | NobodyCam, thanks! g night/morning/afternoon for u too | 20:12 |
| lucasagomes | devananda, g'night! | 20:12 |
| NobodyCam | :) | 20:12 |
| NobodyCam | have a pint for me | 20:12 |
| lucasagomes | NobodyCam, heh I'm looking like a zombie today hah | 20:12 |
| lucasagomes | jetlagged | 20:12 |
| NobodyCam | oh ya.. | 20:12 |
| lucasagomes | :P I will eat something and try to get some sleep early today | 20:12 |
| NobodyCam | :) | 20:12 |
| *** dwalleck has quit IRC | 20:16 | |
| * NobodyCam looks for soem food stuff | 20:19 | |
| *** dwalleck has joined #openstack-ironic | 20:25 | |
| *** dwalleck_ has joined #openstack-ironic | 20:26 | |
| *** hmelevsk has quit IRC | 20:26 | |
| jroll | devananda: replied to your comments on the instance_info patch. if I push a new patchset will it make it difficult for you to read/reply to the comments I made? | 20:29 |
| *** dwalleck has quit IRC | 20:30 | |
| *** mrda_away is now known as mrda | 20:31 | |
| mrda | Morning all | 20:31 |
| jroll | hey mrda | 20:32 |
| mrda | hey jroll | 20:32 |
| *** rloo has quit IRC | 20:33 | |
| *** rloo has joined #openstack-ironic | 20:33 | |
| *** rloo has quit IRC | 20:36 | |
| *** rloo has joined #openstack-ironic | 20:36 | |
| Shrews | adam_g, ifarkas: so, the latest version of ironic needs to be available (with NobodyCam's recent nova driver import stuff) before devstack will work. just 'cd /opt/stack/ironic ; git pull' and it should work again | 20:53 |
| Shrews | i'll update the etherpad and review | 20:54 |
| mrda | Hi Shrews, which etherpad is this? (I might find time to have a play today) | 20:55 |
| Shrews | mrda: https://etherpad.openstack.org/p/IronicAndDevstackAgain | 20:55 |
| mrda | Shrews: ta | 20:55 |
| adam_g | Shrews, whats the nova.conf key=value you're setting? | 20:57 |
| Shrews | adam_g: which setting? | 20:57 |
| adam_g | Shrews, sorry, for the scheduler driver | 20:58 |
| jroll | devananda: replied to your comments on the instance_info patch. if I push a new patchset will it make it difficult for you to read/reply to the comments I made? | 20:58 |
| Shrews | adam_g: i don't set scheduler_driver. scheduler_host_manager (if that's what you meant) is already set correctly | 20:59 |
| Shrews | i did comment out the NOVA_BRANCH setting, though | 21:01 |
| *** vkozhukalov has joined #openstack-ironic | 21:02 | |
| *** jdob has quit IRC | 21:05 | |
| NobodyCam | devananda: happen to be back yet/ | 21:05 |
| *** jbjohnso has quit IRC | 21:10 | |
| NobodyCam | style listic question: does anyone have thoughts on the best way to get the return object/values/whatever from line #149 of https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L145-L161 | 21:11 |
| jroll | NobodyCam: not sure what you mean. what are you trying to do? | 21:15 |
| NobodyCam | jroll: in line 385 of that file that function is called | 21:16 |
| NobodyCam | https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/driver.py#L385 | 21:16 |
| NobodyCam | with node.update | 21:16 |
| NobodyCam | which will return a node object. which I would like to get that object back | 21:16 |
| NobodyCam | one to many which's in that I think | 21:17 |
| jroll | ohhh, how to return something from _retry_if_service_is_unavailable | 21:17 |
| jroll | hrm | 21:17 |
| jroll | you want the node object back in line 385, right? | 21:18 |
| NobodyCam | yeppers | 21:19 |
| *** dwalleck_ has quit IRC | 21:21 | |
| jroll | oh, NobodyCam, you can pass a return value to LoopingCallDone | 21:21 |
| NobodyCam | I was thinking if func() run with out error then set something like self.return_value and just get it from there | 21:21 |
| jroll | and I bet there's a way to get that out | 21:22 |
| *** linggao has quit IRC | 21:22 | |
| jroll | timer.done will be an eventlet.event.Event | 21:22 |
| jroll | and that event gets sent the return value | 21:23 |
| jroll | so you should be able to use that somehow | 21:23 |
| NobodyCam | jroll: humm | 21:24 |
| jroll | I want to link you to the loopingcall source | 21:24 |
| jroll | but I can't find it in github :P | 21:24 |
| NobodyCam | i'll google | 21:25 |
| jroll | or can't find common, specifically | 21:25 |
| jroll | but check out that FixedIntervalLoopingCall thing | 21:25 |
| NobodyCam | :) | 21:26 |
| jroll | oh, even better | 21:26 |
| jroll | .wait() should return it | 21:26 |
| jroll | this is what eventually gets called with the return value passed into LoopingCallDone() | 21:27 |
| jroll | http://eventlet.net/doc/modules/event.html#eventlet.event.Event.send | 21:27 |
| NobodyCam | oh neat | 21:27 |
| jroll | :) | 21:27 |
| *** mdurnosvistov_ has joined #openstack-ironic | 21:30 | |
| openstackgerrit | Jenkins proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/79334 | 21:31 |
| NobodyCam | jroll: that cool stuff but not quite what I was looking for, As I am looking for the patched object the call will return, with out first creating a node object to pss back | 21:34 |
| jroll | NobodyCam: the node would be returned by func(*args), right? | 21:34 |
| NobodyCam | ya | 21:34 |
| jroll | so grab that and pass it into LoopingCallDone() | 21:35 |
| jroll | and it will be returned by wait() in the last line of your function there | 21:35 |
| jroll | does that make sense? | 21:36 |
| NobodyCam | actually yes | 21:37 |
| jroll | sweet :) | 21:37 |
| NobodyCam | jroll: Thank you:) | 21:38 |
| jroll | no problem | 21:38 |
| *** dwalleck has joined #openstack-ironic | 21:41 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add a JSON type for nested JSON structures https://review.openstack.org/79465 | 21:46 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add Node.instance_info field https://review.openstack.org/79466 | 21:46 |
| jroll | ugh, that got screwed up | 21:48 |
| NobodyCam | i distracted you | 21:49 |
| openstackgerrit | Ruby Loo proposed a change to openstack/python-ironicclient: Add driver-properties command https://review.openstack.org/76338 | 21:49 |
| jroll | NobodyCam: nah, it was a weird thing where I wanted to split something off to a separate commit before the commit in review :/ | 21:50 |
| jroll | idk if it was avoidable | 21:50 |
| NobodyCam | :) | 21:51 |
| *** max_lobur has quit IRC | 21:53 | |
| *** max_lobur has joined #openstack-ironic | 21:55 | |
| *** vkozhukalov has quit IRC | 21:58 | |
| *** dwalleck_ has joined #openstack-ironic | 21:58 | |
| *** dwalleck has quit IRC | 22:01 | |
| rloo | russell_h: did you get an answer to database.connection? | 22:02 |
| *** dwalleck has joined #openstack-ironic | 22:03 | |
| *** dwalleck_ has quit IRC | 22:03 | |
| russell_h | rloo: nope | 22:06 |
| NobodyCam | humm I am having a issue installing glance-client | 22:06 |
| russell_h | rloo: I'm not stuck on it, was just curious if it was intended | 22:06 |
| *** vkozhukalov has joined #openstack-ironic | 22:06 | |
| rloo | russell_h: i just encountered it, but I have to take off. I tracked it down to this change: https://github.com/openstack/ironic/commit/36a9682478062b099456dfb7586fc524a435ded5. | 22:07 |
| rloo | russell_h: If I get a chance, I'll take a look tonight. | 22:07 |
| NobodyCam | heading back north rloo ? | 22:08 |
| *** romcheg1 has quit IRC | 22:12 | |
| devananda | NobodyCam: back now | 22:13 |
| NobodyCam | wb devananda | 22:14 |
| NobodyCam | I jroll and I talked out my issue | 22:14 |
| NobodyCam | lol tring to get a test env going now | 22:14 |
| devananda | :) | 22:15 |
| jroll | devananda: I replied to your comments here: https://review.openstack.org/#/c/79396/ | 22:16 |
| jroll | devananda: but when I pulled the one thing out into a new commit it ended up making new reviews: https://review.openstack.org/#/c/79465/ https://review.openstack.org/#/c/79466/ | 22:17 |
| devananda | jroll: https://wiki.openstack.org/wiki/GerritWorkflow | 22:18 |
| jroll | devananda: what are you pointing to there? | 22:19 |
| jroll | devananda: you wanted it in a separate review, and I wanted my main changes to depend on that commit | 22:19 |
| jroll | and I think because I added a new commit *before* the commit under review, that's why it made a new one | 22:19 |
| devananda | jroll: it didn't have to create a new patch. you seemed to indicate that it was an accident | 22:19 |
| devananda | jroll: nope, that's not why | 22:19 |
| jroll | hm | 22:20 |
| jroll | do I just need to manually add the change id or something? | 22:20 |
| devananda | jroll: gerrit uses the Change-Id line | 22:20 |
| devananda | to tell which patch set is which | 22:20 |
| devananda | not the commit sha | 22:20 |
| devananda | so yes | 22:20 |
| jroll | ah right | 22:20 |
| devananda | or do a rebase -i | 22:20 |
| jroll | ok, let me amend the latter review | 22:20 |
| devananda | it's fine | 22:20 |
| jroll | ok | 22:21 |
| jroll | thank you | 22:21 |
| devananda | jroll: it looks like the two new reviews supercede the first one. if so, and you don't intend to continue it, please abandon it | 22:22 |
| devananda | jroll: i'm also curious why you need a nested dict here | 22:22 |
| jroll | devananda: sure. just wanted to make sure you saw the comments before I did so | 22:22 |
| devananda | ack, thanks | 22:22 |
| jroll | devananda: see the blueprint | 22:22 |
| jroll | devananda: unless you think a bunch of keys like "configdrive__ssh_keys", "configdrive__admin_pass", etc. are better | 22:23 |
| *** dwalleck has quit IRC | 22:23 | |
| devananda | ahh | 22:24 |
| *** max_lobur has quit IRC | 22:26 | |
| devananda | I think that is compatible with the data structure we have today, whereby if a parameter is present, it is assumed to be a literal value (eg, not a dict) | 22:26 |
| devananda | adding nested and unversioned dicts will make maintaining compatibility across upgrades more difficult | 22:27 |
| devananda | projects have been trying to solve this internally by versioning their RPC and DB objects (rather than just passing dicts and nested dicts around) | 22:27 |
| devananda | that is generally solved between projects by versioning the API | 22:27 |
| jroll | yeah, though it's hard to version something that we are not doing validation on :/ | 22:28 |
| devananda | we've already made the API somewhat flexible at specific places, eg. driver_info, because different drivers requrie different info | 22:28 |
| devananda | jroll: we do validate | 22:28 |
| devananda | jroll: each driver must expose a validate() method | 22:28 |
| devananda | precisely because of this | 22:29 |
| jroll | devananda: hmm, ok. | 22:29 |
| jroll | I'm fine with versioning it, but I'm not sure the best way to do that (field in the JSON or an extra field in the DB) | 22:29 |
| devananda | i'm looking into the docs a bit | 22:33 |
| devananda | http://docs.openstack.org/user-guide/content/config_reference.html is interesting | 22:33 |
| jroll | that's not really related with what goes in the configdrive | 22:33 |
| devananda | it's related to what nova will need to instruct Irnic to do | 22:34 |
| devananda | which needs to be part of the API to support config drive in Ironic | 22:34 |
| devananda | eg, config_drive_cdrom | 22:35 |
| devananda | config_drive_inject_password | 22:35 |
| devananda | etc | 22:35 |
| jroll | hmm. | 22:35 |
| jroll | yeah | 22:35 |
| devananda | other drivers can mount a CDROM using the OOB channel | 22:36 |
| devananda | which is a very compelling use case | 22:36 |
| jroll | we were planning in the first version to only support a subset of that | 22:36 |
| jroll | but yeah, I see where you're going | 22:36 |
| devananda | as it can deliver secure metadata | 22:36 |
| jroll | yeah | 22:36 |
| devananda | what's the max size of a config drive? | 22:36 |
| devananda | i'm looking into nova's code -- https://github.com/openstack/nova/blob/master/nova/virt/configdrive.py | 22:37 |
| jroll | I'm not sure there is one | 22:37 |
| jroll | we're using 16MB as the partition size for it right now | 22:37 |
| devananda | a JSONEncodedDict, whether varchar or TEXT, is not suitable to store the whole contents of the config drive | 22:37 |
| jroll | but you can basically inject arbitrary files into the system through configdrive | 22:37 |
| jroll | why? | 22:37 |
| devananda | if the only thing passed in is an SSH pub key and an admin pass -- sure, it would work | 22:37 |
| devananda | databases are terrible places to store large files | 22:38 |
| jroll | because you don't want to parse a massive blob of json? | 22:38 |
| jroll | right | 22:38 |
| jroll | so, we had another idea with using swift | 22:38 |
| devananda | ++ | 22:38 |
| jroll | and storing a pointer in the db | 22:38 |
| devananda | much better | 22:38 |
| jroll | props to JoshNang on that one | 22:38 |
| devananda | let nova build the config drive, pass options and a swift ref | 22:38 |
| devananda | or a ref to $somewhere | 22:38 |
| JoshNang | :) | 22:39 |
| jroll | well | 22:39 |
| jroll | I mostly agree | 22:39 |
| JayF | Are we sure we want the configdrive to be opaque to Ironic? | 22:39 |
| jroll | although I'd almost rather pass the metadata around in json and have the agent format it and write it | 22:39 |
| jroll | rather than nova passing us an opaque iso image | 22:40 |
| JoshNang | JayF: is there a benefit to ironic being able to look inside the configdrive? | 22:40 |
| *** matty_dubs is now known as matty_dubs|gone | 22:40 | |
| devananda | nova already has the options and the code to build both vfat and iso configdrive images | 22:40 |
| jroll | right, I know | 22:40 |
| JayF | JoshNang: hmm. I'm not sure, I'm trying to think it through myself still. | 22:40 |
| *** dwalleck has joined #openstack-ironic | 22:41 | |
| devananda | i can see a case where particular ironic drivers may need to build an ISO and mount it OOB -- eg, for firmware update, if that's the vendor's chosen channel | 22:41 |
| JayF | For instance, are there network configurations specific to physical devices (like bonded interfaces or trunked vlans) that nova might not support | 22:41 |
| devananda | JayF: if so, we should add support to nova | 22:42 |
| devananda | JayF: or neutron | 22:42 |
| JayF | Does ironic already depend on direct access to swift? i.e. not via glance? | 22:45 |
| russell_h | JoshNang: Ironic may have to look in or modify things in configdrive | 22:48 |
| russell_h | JoshNang: for example, network configurations | 22:48 |
| NobodyCam | JayF: I do not recall ironic depending on swift access | 22:48 |
| russell_h | JoshNang: at least in our case - ideally eventually DHCP, etc, but.. :) | 22:48 |
| JayF | So storing configdrive information in swift would add that dependency for the driver. IDK if that's a big deal or not, but it's worth being explicitly said | 22:49 |
| *** dwalleck has quit IRC | 22:49 | |
| devananda | russell_h: if someone is using cloud-init, that's not possible | 22:49 |
| russell_h | whats not? | 22:49 |
| *** dwalleck has joined #openstack-ironic | 22:49 | |
| devananda | russell_h: changing the config drive | 22:50 |
| devananda | russell_h: unless ironic talks back to nova to change the metadata | 22:50 |
| NobodyCam | JayF: ++ for explicitly calling out! | 22:50 |
| JayF | I think he means in flight, not once written | 22:50 |
| * NobodyCam tihnks we may want a driver-requirments.txt file | 22:50 | |
| devananda | russell_h: what part of network configuration couldn't be supplied to nova / neutron and passed into the instance that way? | 22:52 |
| *** bearhands is now known as comstud | 22:52 | |
| NobodyCam | devananda: what are your thoughts on somehting like a single driver adding a dependicy to requirments.txt file? | 22:52 |
| devananda | russell_h: ironic is already passing node MAC addresses up the stack so they are correctly configured in nova-network / neutron | 22:52 |
| devananda | NobodyCam: well, that's already been done, but actually should not have been | 22:52 |
| NobodyCam | done for SM? | 22:53 |
| devananda | NobodyCam: we should remove python-seamicroclient from requierments.txt and add it as a dynamic load in the driver | 22:53 |
| devananda | ya | 22:53 |
| NobodyCam | ya | 22:53 |
| russell_h | devananda: last week when we were talking, didn't we conclude that we should talk to neutron directly from within our driver (or at least in Ironic) instead of trying to have nova do that? | 22:54 |
| devananda | russell_h: when needing to make changes to information in neutron, yes | 22:56 |
| devananda | russell_h: AIUI, generating the instance metadata is nova's responsibility. not ironic's. we still need to expose a means for nova to do that | 22:57 |
| devananda | regardless of how that metadata is delivered | 22:57 |
| *** derekh has joined #openstack-ironic | 22:57 | |
| devananda | so if nova needs certain info (eg, MAC addresses, port bonding info, etc) in order to generate the instance's meta data, ironic needs to expose that back | 22:58 |
| russell_h | that feels broken | 22:58 |
| russell_h | well, hmm | 22:58 |
| devananda | lemme check with nova on something | 22:59 |
| *** dwalleck has quit IRC | 23:00 | |
| *** martyntaylor has left #openstack-ironic | 23:00 | |
| *** mdurnosvistov_ has quit IRC | 23:01 | |
| devananda | russell_h: answer from nova seems to be "there's probably no way, and if there is, it's against policy" | 23:03 |
| NobodyCam | brb quick walkies | 23:04 |
| russell_h | devananda: this seems odd. Shouldn't a hypervisor have the freedom to pass configuration information into an instance? | 23:04 |
| russell_h | also, comstud was telling me about a thing they added in nova to allow Neutron to pass back information | 23:05 |
| russell_h | for a somewhat different reason | 23:05 |
| russell_h | not sure of the details, but the idea, as I understood it, was that Neutron would update nova when things changed | 23:05 |
| russell_h | comstud: ^ | 23:05 |
| comstud | heh, was just talking in -nova | 23:06 |
| comstud | hm | 23:06 |
| comstud | reading back a bit | 23:06 |
| comstud | as far as the neturon thing you're referring to.. we have implemented a general event callback mechansim | 23:07 |
| comstud | 2 use cases for neutron 1) to tell when port attachment finished, 2) to tell when something changed so nova can update its cache | 23:08 |
| comstud | there's a use case for cinder.. not sure if that's appropriate for ironic in any way | 23:08 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add Node.instance_info field https://review.openstack.org/79466 | 23:10 |
| NobodyCam | anyone else having issue with python-glanceclient pip installs? | 23:13 |
| devananda | comstud: yea, it's appropriate as well | 23:16 |
| devananda | comstud: eg, to update nova if ironic notices the node's power state has changed | 23:17 |
| devananda | comstud: or to update nova/neutron if the operator replaces a faulty NIC in a baremetal instance without taking the isntance offline | 23:17 |
| comstud | ah yeah | 23:19 |
| comstud | although the latter would happen from compute's periodic task | 23:19 |
| comstud | er | 23:19 |
| comstud | former | 23:20 |
| comstud | the latter maybe also... but i could see nova being notified if that doesn't work, yeah | 23:20 |
| rloo | NobodyCam: got back Saturday. Nice to be home again (although when I was chipping the ice today I wondered...) | 23:22 |
| NobodyCam | rloo: hehehehe :) | 23:22 |
| NobodyCam | rloo: http://wattsupwiththat.files.wordpress.com/2011/02/al-gore-flamethrower1.jpg | 23:25 |
| NobodyCam | lol beets chipping | 23:26 |
| rloo | NobodyCam: cool. I'll check ebay to see if I can get one cheap :-) | 23:26 |
| NobodyCam | lol | 23:26 |
| devananda | comstud: yea, power state sync via periodic is probably fine | 23:26 |
| jroll | devananda: so uh, are you still -1 on the agent writing the configdrive? | 23:34 |
| JayF | FWIW; this isn't theoretical. The agent /already/ writes a configdrive, right? | 23:35 |
| devananda | jroll: the agent *copying* the config drive from somewhere -- that's fine | 23:35 |
| jroll | JayF: correct | 23:35 |
| jroll | devananda: so you want it to be an opaque blob? | 23:35 |
| devananda | jroll: if it's not, one of two things happen -- we copy the nova code that builds a config drive, or we do it differently and are incompatible with other hypervisors | 23:36 |
| devananda | jroll: both are non-starters IMO. unless you see other options, I think the agent needs to treat it as a raw disk | 23:37 |
| jroll | we're *already* doing it in a compatible manner, although probably not as flexible | 23:37 |
| comstud | what I'd really like to do is just have nova pass in "here's the config drive" | 23:49 |
| comstud | because that's really how all of the virt drivers work | 23:49 |
| comstud | (although instead of it being passed in to them, they actually just call the same library call and get back an image to write out) | 23:49 |
| comstud | but i think russell_h doesn't like it just being a base64'd image passed in? | 23:51 |
| openstackgerrit | Adam Gandelman proposed a change to openstack/python-ironicclient: Avoid traceback with insufficient auth credentials https://review.openstack.org/79494 | 23:52 |
| jroll | comstud: we feel like there could be a need to modify it in ironic, e.g. for networking configuration | 23:52 |
| jroll | or injecting an extra file | 23:52 |
| comstud | yeah | 23:52 |
| russell_h | comstud: I mean, I don't like the base64, but thats because I think the virt driver is the wrong abstraction for this | 23:53 |
| jroll | (thinking third party drivers here) | 23:53 |
| comstud | i'd rather nova knew to do it | 23:53 |
| comstud | but | 23:53 |
| comstud | I know | 23:53 |
| comstud | russell_h: Yeah, maybe that's the real problem | 23:53 |
| russell_h | comstud: but if I accept the virt driver API, my concern is just about injecting configuration into the configdrive if its a base64 blob | 23:53 |
| JayF | The thing I don't get, is why it's bad to pass around what the configdrive should look like, as an object, then write it out ourselves. That seems like it lends itself to more delivery mechanisms and injection if needed along the way | 23:53 |
| comstud | We can pass in the directory layout and contents that nova knows about, instead of the fat or iso image | 23:54 |
| comstud | but we'll need to expose that in the configdrive builder module/library | 23:54 |
| comstud | because right now everything just asks for the fat or iso or ext*fs image or whatever | 23:54 |
| russell_h | I mean, I'm not actually concerned about duplicating the configdrive builder thing | 23:55 |
| russell_h | its not much code | 23:55 |
| comstud | the image part isn't | 23:55 |
| comstud | we can bypass that easily and get the list of files/contents | 23:55 |
| russell_h | ah | 23:56 |
| comstud | What actually generates the dictionary of files/contents is actually a mess | 23:56 |
| russell_h | why? | 23:56 |
| comstud | ugly code, imo :) | 23:56 |
| comstud | it's all in nova/api/metadata/base.py | 23:56 |
| comstud | the code organization is... not right | 23:57 |
| russell_h | ah | 23:57 |
| comstud | config drive builder is using code stuffed into the metadata service | 23:57 |
| comstud | it should be pulled out into its own lib | 23:57 |
| comstud | and shared metween md service and config drive | 23:57 |
| russell_h | ah, yeah this is ugly | 23:57 |
| comstud | between | 23:57 |
| comstud | I wanted to barf when I went to look at this a month ago or more when i was talking with jroll | 23:58 |
| comstud | anyway, config drive builder just calls into that base.py file | 23:58 |
| comstud | ironic driver could do the same and just pass the invidiual files and ironic could add things and build image itself | 23:59 |
| russell_h | +1 | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
