| *** marcoemorais has quit IRC | 00:00 | |
| *** marcoemorais has joined #openstack-ironic | 00:00 | |
| NobodyCam | IPA team: are lines 12 thru 17 of https://review.openstack.org/#/c/103105/5/elements/ironic-agent/post-install.d/80-ironic-agent required for IPA? | 00:09 |
|---|---|---|
| *** lucas-dinner has quit IRC | 00:17 | |
| jroll | NobodyCam: idk, I think that's fedora related maybe? | 00:18 |
| jroll | NobodyCam: no idea, at all | 00:19 |
| NobodyCam | seems wrong to me | 00:20 |
| jroll | yeah, dunno, never seen anything like that | 00:21 |
| NobodyCam | hat looks like the old school file injection from early nova baremetal | 00:22 |
| NobodyCam | s/hat/that/ | 00:22 |
| jroll | gah | 00:24 |
| *** nosnos has joined #openstack-ironic | 00:24 | |
| *** ChuckC_ has joined #openstack-ironic | 00:29 | |
| NobodyCam | yuriyz: please ping me RE: https://review.openstack.org/#/c/103105 | 00:32 |
| *** ChuckC has joined #openstack-ironic | 00:33 | |
| *** ChuckC_ has quit IRC | 00:33 | |
| *** todd_dsm has joined #openstack-ironic | 00:39 | |
| *** Nicklaus_Zulauf3 has quit IRC | 00:42 | |
| *** Arvid_Ebert86 has joined #openstack-ironic | 00:42 | |
| *** penick has joined #openstack-ironic | 00:46 | |
| *** marcoemorais has quit IRC | 00:47 | |
| *** marcoemorais has joined #openstack-ironic | 00:48 | |
| *** marcoemorais has quit IRC | 00:48 | |
| *** marcoemorais has joined #openstack-ironic | 00:48 | |
| *** marcoemorais has quit IRC | 00:48 | |
| *** marcoemorais has joined #openstack-ironic | 00:49 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Use LLDP to get switch port mapping https://review.openstack.org/92627 | 00:50 |
| *** r-daneel_ has quit IRC | 00:56 | |
| *** todd_dsm has quit IRC | 01:02 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Use LLDP to get switch port mapping https://review.openstack.org/92627 | 01:08 |
| *** chenglch has joined #openstack-ironic | 01:18 | |
| *** todd_dsm has joined #openstack-ironic | 01:20 | |
| *** chenglch has quit IRC | 01:20 | |
| *** Arvid_Ebert86 has quit IRC | 01:25 | |
| *** chenglch has joined #openstack-ironic | 01:29 | |
| openstackgerrit | xu-haiwei proposed a change to openstack/ironic: Fix typo to show correct log message https://review.openstack.org/122316 | 01:34 |
| *** Haomeng has joined #openstack-ironic | 01:39 | |
| *** Haomeng|2 has quit IRC | 01:40 | |
| *** todd_dsm has quit IRC | 01:42 | |
| *** Alyson_Cassin83 has joined #openstack-ironic | 01:44 | |
| *** todd_dsm has joined #openstack-ironic | 02:01 | |
| *** todd_dsm has quit IRC | 02:02 | |
| *** marcoemorais has quit IRC | 02:02 | |
| *** pcrews has quit IRC | 02:06 | |
| *** todd_dsm has joined #openstack-ironic | 02:11 | |
| *** Poornima has joined #openstack-ironic | 02:14 | |
| openstackgerrit | xu-haiwei proposed a change to openstack/ironic: Sync ironic/openstack/common/policy.py with oslo. https://review.openstack.org/122316 | 02:15 |
| *** dlaube has quit IRC | 02:19 | |
| *** todd_dsm has quit IRC | 02:24 | |
| *** harlowja is now known as harlowja_away | 02:36 | |
| *** Alyson_Cassin83 has quit IRC | 02:38 | |
| *** Alysa_Schmeler has joined #openstack-ironic | 02:49 | |
| *** ramineni has joined #openstack-ironic | 02:55 | |
| *** sabah has joined #openstack-ironic | 02:58 | |
| *** comstud has quit IRC | 03:07 | |
| *** wendar has quit IRC | 03:07 | |
| *** wendar has joined #openstack-ironic | 03:08 | |
| JayF | jroll: is there a bug number yet for the new agent tempest job failures? | 03:15 |
| *** penick has quit IRC | 03:19 | |
| *** jrist has quit IRC | 03:33 | |
| *** Alysa_Schmeler has quit IRC | 03:39 | |
| *** jrist has joined #openstack-ironic | 03:40 | |
| openstackgerrit | Yuiko Takada proposed a change to openstack/python-ironicclient: Add unit test for "ironic node-update" shell cmd https://review.openstack.org/120989 | 03:41 |
| *** Albert87 has joined #openstack-ironic | 03:43 | |
| *** marcoemorais has joined #openstack-ironic | 03:55 | |
| *** sabah has quit IRC | 03:56 | |
| *** Poornima has quit IRC | 03:56 | |
| *** marcoemorais1 has joined #openstack-ironic | 03:57 | |
| *** lazy_prince is now known as killer_prince | 04:00 | |
| *** marcoemorais has quit IRC | 04:00 | |
| *** Haomeng|2 has joined #openstack-ironic | 04:07 | |
| *** Haomeng has quit IRC | 04:08 | |
| openstackgerrit | Yuiko Takada proposed a change to openstack/python-ironicclient: Add unit tests for "ironic node-show" shell cmd https://review.openstack.org/120001 | 04:08 |
| openstackgerrit | Yuiko Takada proposed a change to openstack/python-ironicclient: Add unit tests for "ironic node-create" shell cmd https://review.openstack.org/120312 | 04:15 |
| *** Albert87 has quit IRC | 04:18 | |
| *** comstud has joined #openstack-ironic | 04:21 | |
| *** rushiagr_away is now known as rushiagr | 04:23 | |
| *** nikunj2512 has joined #openstack-ironic | 04:27 | |
| *** Poornima has joined #openstack-ironic | 04:33 | |
| *** comstud has quit IRC | 04:35 | |
| *** killer_prince has quit IRC | 04:36 | |
| *** comstud has joined #openstack-ironic | 04:38 | |
| *** vinbs has joined #openstack-ironic | 04:40 | |
| *** rakesh_hs has joined #openstack-ironic | 04:41 | |
| *** comstud has quit IRC | 04:45 | |
| *** pensu has joined #openstack-ironic | 05:00 | |
| *** r-mibu has joined #openstack-ironic | 05:10 | |
| *** comstud has joined #openstack-ironic | 05:18 | |
| *** pensu has quit IRC | 05:19 | |
| *** pensu has joined #openstack-ironic | 05:23 | |
| *** pensu has quit IRC | 06:00 | |
| *** jcoufal has joined #openstack-ironic | 06:01 | |
| *** killer_prince has joined #openstack-ironic | 06:07 | |
| *** killer_prince is now known as lazy_prince | 06:07 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/123342 | 06:07 |
| *** k4n0 has joined #openstack-ironic | 06:09 | |
| *** pensu has joined #openstack-ironic | 06:21 | |
| *** athomas has joined #openstack-ironic | 06:43 | |
| *** jistr has joined #openstack-ironic | 06:46 | |
| *** romcheg has joined #openstack-ironic | 07:00 | |
| *** ifarkas has joined #openstack-ironic | 07:02 | |
| *** Poornima has quit IRC | 07:03 | |
| *** Poornima has joined #openstack-ironic | 07:10 | |
| *** GheRivero_ has joined #openstack-ironic | 07:11 | |
| *** GheRivero has quit IRC | 07:14 | |
| *** GheRivero_ is now known as GheRivero | 07:15 | |
| *** romcheg has joined #openstack-ironic | 07:18 | |
| *** romcheg has left #openstack-ironic | 07:33 | |
| *** romcheg has joined #openstack-ironic | 07:33 | |
| *** ndipanov_gone is now known as ndipanov | 07:35 | |
| *** Poornima has quit IRC | 07:44 | |
| *** athomas has quit IRC | 07:51 | |
| *** athomas has joined #openstack-ironic | 07:54 | |
| *** takadayuiko has joined #openstack-ironic | 08:05 | |
| *** lucasagomes has joined #openstack-ironic | 08:17 | |
| *** dtantsur|afk is now known as dtantsur | 08:25 | |
| dtantsur | Morning Ironic | 08:25 |
| vinbs | Morning dtantsur! | 08:25 |
| *** krykowski has joined #openstack-ironic | 08:26 | |
| *** MattMan has quit IRC | 08:32 | |
| *** MattMan has joined #openstack-ironic | 08:33 | |
| *** Poornima has joined #openstack-ironic | 08:40 | |
| *** marcoemorais1 has quit IRC | 08:41 | |
| yuriyz | morning Ironic dtantsur vinbs | 08:41 |
| dtantsur | hi yuriyz, vinbs! :) | 08:42 |
| *** bluex-pl has joined #openstack-ironic | 08:42 | |
| *** takadayuiko has quit IRC | 08:42 | |
| openstackgerrit | Syed Ismail Faizan Barmawer proposed a change to openstack/ironic: pxe_ilo driver to call iLO set_boot_device https://review.openstack.org/123660 | 08:45 |
| *** jistr has quit IRC | 08:45 | |
| vinbs | Morning yuriyz | 08:46 |
| *** christopheraedo has quit IRC | 08:47 | |
| *** christopheraedo has joined #openstack-ironic | 08:48 | |
| openstackgerrit | Syed Ismail Faizan Barmawer proposed a change to openstack/ironic: pxe_ilo driver to call iLO set_boot_device https://review.openstack.org/122819 | 08:48 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: EXPERIMENTAL Implement hardware discovery in PXE driver https://review.openstack.org/110031 | 08:53 |
| yuriyz | 122819 and 123660 duplicated, one should be abandoned | 08:53 |
| *** bluex-pl has quit IRC | 08:53 | |
| *** bluex-pl has joined #openstack-ironic | 08:54 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/123662 | 08:57 |
| *** krykowski has quit IRC | 08:58 | |
| *** krykowski has joined #openstack-ironic | 08:59 | |
| *** bluex-pl has quit IRC | 08:59 | |
| *** bluex-pl has joined #openstack-ironic | 09:00 | |
| *** Haomeng has joined #openstack-ironic | 09:00 | |
| pensu | Hi all, I have set up Ironic using devstack. Now, I need to do some experiments by adding a new node. I couldn't find any guide on how to add a new node to this setup. How can it be done? | 09:00 |
| *** Haomeng|2 has quit IRC | 09:01 | |
| lazy_prince | pensu: are you looking for ironic node-create .... ? | 09:01 |
| dtantsur | pensu, simple example is in the end of http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#exercising-the-services-locally | 09:02 |
| pensu | lazy_prince: yeah, what I want to know is can I add a physical node using that in a devstack rnvironment? I tried and it says can't validate PXE bootloader..... | 09:06 |
| *** krykowski has quit IRC | 09:06 | |
| lazy_prince | pense: aha.. what driver are you using when you are trying to add node..? | 09:07 |
| pensu | dtantsur: yeah, I tried that, just want to confirm that it would work with devstack installation or not? | 09:07 |
| dtantsur | pensu, well, difference is driver used (pxe_ssh) and set of required parameters. And for real bare metal you probably want pxe_ipmitool or another real driver | 09:08 |
| dtantsur | pensu, what exactly says "can't validate PXE bootloader"? which command? | 09:08 |
| *** jistr has joined #openstack-ironic | 09:08 | |
| dtantsur | usually pxe parameters are filled by nova when deploying a node | 09:08 |
| dtantsur | and ssh/ipmi/ilo/drac ones should be set by you in advance | 09:09 |
| *** derekh has joined #openstack-ironic | 09:09 | |
| pensu | lazy_prince: dtantsur: my nvoa.conf says "compute_driver = ironic.nova.virt.ironic.IronicDriver" | 09:09 |
| pensu | does that makes sense? | 09:10 |
| dtantsur | pensu, I mean Ironic driver that is set for every node, not Nova compute driver | 09:10 |
| dtantsur | do ironic node-show for any existing node and you'll see | 09:10 |
| *** Nisha has joined #openstack-ironic | 09:10 | |
| pensu | dtantsur: okay, I guess I understand this. Let me try something and I will get back to you. Thank you....:) | 09:13 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Remove redundant context parameter https://review.openstack.org/123412 | 09:14 |
| lucasagomes | dtantsur, can you re-approve it please? https://review.openstack.org/#/c/123412/ | 09:16 |
| lucasagomes | I had to rebase because the dependent patch was outdated (now merged) | 09:16 |
| *** andreykurilin_ has joined #openstack-ironic | 09:16 | |
| openstackgerrit | A change was merged to openstack/ironic: Update docs with new dbsync command https://review.openstack.org/122802 | 09:16 |
| dtantsur | lucasagomes, done | 09:19 |
| lucasagomes | dtantsur, ta much | 09:19 |
| *** krykowski has joined #openstack-ironic | 09:20 | |
| *** loki184 has joined #openstack-ironic | 09:20 | |
| *** openstack has joined #openstack-ironic | 09:24 | |
| zigo | devananda: When I see that the global-requirements.txt pushes me to do a new version, I just package last tag. Currently 0.2.1 is in Debian Experimental. Do you think I should package something more recent? | 09:25 |
| *** vdrok has quit IRC | 09:32 | |
| *** vdrok has joined #openstack-ironic | 09:36 | |
| *** Nisha has quit IRC | 09:40 | |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Do not return 'id' in REST API error messages https://review.openstack.org/123150 | 09:45 |
| ramineni | lucasagomes:hi | 09:46 |
| lucasagomes | ramineni, hi there | 09:47 |
| ramineni | lucasagomes: regarding review comment on https://review.openstack.org/#/c/122819/4/ironic/drivers/modules/ilo/deploy.py | 09:47 |
| ramineni | raising IPMIFailure in set_boot_device | 09:48 |
| ramineni | lucasagomes: we are not using ipmi call to set the boot device , so doesn't it convey the wrong idea , if we raise IPMIfailure | 09:49 |
| lucasagomes | ramineni, hmm... to be honest yeah, but I was thinking about the user perspective | 09:49 |
| lucasagomes | like that interface inherits from the ipmitool one | 09:49 |
| lucasagomes | so for get_boot_device for e.g | 09:49 |
| lucasagomes | if something goes wrong, IPMIFailure will be raised | 09:49 |
| lucasagomes | same for get_sensor_data | 09:50 |
| lucasagomes | now for set_boot_device IloOperationError will be raised | 09:50 |
| ramineni | lucasagomes: yes , for others i agree , but if we keep it here , user will get worng idea , something wrong with IPMI? | 09:50 |
| ramineni | yes , it raises IloOperationError | 09:50 |
| ramineni | for now | 09:51 |
| lucasagomes | ramineni, right, ok well it was mostly a suggestion. But I can see benefits on both approachs | 09:51 |
| lucasagomes | but yeah I'm grand raising iloOperationError if you prefer. But we still need to updated that docs about that | 09:51 |
| ramineni | lucasagomes: yes ..will do that | 09:52 |
| lucasagomes | ramineni, and, there's a duplicated patch. It would be good to abandon one of them | 09:52 |
| lucasagomes | (they look exactly same) | 09:52 |
| ramineni | lucasagomes: ya .. didn't check Change-Id has been changed :( | 09:52 |
| ramineni | will abandon that | 09:52 |
| lucasagomes | oh, no worries :) | 09:52 |
| lucasagomes | alright so we can keep the ilo error | 09:53 |
| ramineni | lucasagomes: ya ..thanks :) | 09:53 |
| *** Nisha has joined #openstack-ironic | 09:54 | |
| *** viktors|afk is now known as viktors | 09:56 | |
| openstackgerrit | Syed Ismail Faizan Barmawer proposed a change to openstack/ironic: pxe_ilo driver to call iLO set_boot_device https://review.openstack.org/122819 | 09:58 |
| *** krykowski has quit IRC | 10:01 | |
| dtantsur | btw I'd also prefer iLO error | 10:03 |
| openstackgerrit | A change was merged to openstack/ironic: Update devstack docs, require Ubuntu 14.04 https://review.openstack.org/122519 | 10:04 |
| *** andreykurilin_ has quit IRC | 10:28 | |
| *** nikunj2512 has quit IRC | 10:30 | |
| *** datajerk has joined #openstack-ironic | 10:35 | |
| *** krykowski has joined #openstack-ironic | 10:36 | |
| *** rushiagr is now known as rushiagr_away | 10:40 | |
| *** rushiagr_away is now known as rushiagr | 10:43 | |
| *** pensu has quit IRC | 10:45 | |
| *** bluex-pl has quit IRC | 10:56 | |
| *** bluex-pl has joined #openstack-ironic | 10:56 | |
| openstackgerrit | A change was merged to openstack/ironic: Remove redundant context parameter https://review.openstack.org/123412 | 10:57 |
| *** ramineni has quit IRC | 11:02 | |
| *** krykowski has quit IRC | 11:06 | |
| *** romcheg has quit IRC | 11:08 | |
| *** lucasagomes is now known as lucas-hungry | 11:09 | |
| *** chenglch has quit IRC | 11:15 | |
| ekarlso | Hey guys, is ironic using WSME or JsonSchema ? | 11:21 |
| *** vinbs has quit IRC | 11:23 | |
| dtantsur | ekarlso, wsme at least | 11:27 |
| ekarlso | dtantsur: using both or ? | 11:28 |
| dtantsur | not sure | 11:29 |
| *** bluex-pl has quit IRC | 11:30 | |
| *** bluex-pl has joined #openstack-ironic | 11:31 | |
| *** krykowski has joined #openstack-ironic | 11:33 | |
| *** pensu has joined #openstack-ironic | 11:34 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/123662 | 11:34 |
| *** bluex-pl has quit IRC | 11:35 | |
| *** bluex-pl has joined #openstack-ironic | 11:36 | |
| *** Nisha has quit IRC | 11:42 | |
| *** Poornima has quit IRC | 11:55 | |
| *** romcheg has joined #openstack-ironic | 11:58 | |
| *** loki184 has quit IRC | 12:04 | |
| Shrews | morning ironic | 12:04 |
| Shrews | ekarlso: the ironic api server uses wsme for validation | 12:05 |
| ekarlso | Shrews: cool! :) | 12:05 |
| *** lucas-hungry is now known as lucasagomes | 12:08 | |
| lucasagomes | Shrews, morning | 12:08 |
| Shrews | lucasagomes: morning | 12:09 |
| *** HenryG_afk is now known as HenryG | 12:09 | |
| *** krykowski_ has joined #openstack-ironic | 12:09 | |
| *** krykowski has quit IRC | 12:13 | |
| *** pensu has quit IRC | 12:15 | |
| *** bluex-pl has quit IRC | 12:16 | |
| *** bluex-pl has joined #openstack-ironic | 12:22 | |
| *** linggao has joined #openstack-ironic | 12:27 | |
| *** bluex-pl has quit IRC | 12:29 | |
| *** krykowski_ has quit IRC | 12:30 | |
| *** pensu has joined #openstack-ironic | 12:30 | |
| *** bluex-pl has joined #openstack-ironic | 12:31 | |
| *** rakesh_hs has quit IRC | 12:36 | |
| jroll | JayF: no | 12:40 |
| jroll | morning ironic :) | 12:40 |
| jroll | pensu: perhaps you could show us the node-create command you are running, and the exact error? | 12:42 |
| lucasagomes | jroll, morning | 12:43 |
| *** bluex-pl has quit IRC | 12:46 | |
| ekarlso | Shrews: was there a reason why WSME vs jsonschema ? | 12:50 |
| Shrews | ekarlso: before my time, but i think that code was pulled from another project to start with? also, the author of wsme works on openstack (dhellman) | 12:51 |
| lucasagomes | ekarlso, it was recommended to use WSME for new projects | 12:51 |
| lucasagomes | yeah the code was based on ceilometers api | 12:51 |
| Shrews | ekarlso: also, wsme was used b/c at one point, openstack supported both json AND xml, which wsme supports quite nicely | 12:52 |
| *** jjohnson2 has joined #openstack-ironic | 12:57 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: EXPERIMENTAL Implement hardware discovery in PXE driver https://review.openstack.org/110031 | 12:58 |
| ekarlso | ah ok | 12:59 |
| dtantsur | Shrews, jroll, morning | 12:59 |
| Shrews | dtantsur: o/ | 12:59 |
| jroll | goooood morning :) | 13:00 |
| * jroll really wants read-only access via policy.conf right now | 13:00 | |
| *** rloo has joined #openstack-ironic | 13:02 | |
| openstackgerrit | A change was merged to openstack/ironic: pxe_ilo driver to call iLO set_boot_device https://review.openstack.org/122819 | 13:03 |
| *** vdrok has quit IRC | 13:08 | |
| dtantsur | jroll, I guess we can update whiteboard re the mirrors? | 13:10 |
| rloo | hi everyone! | 13:11 |
| lucasagomes | rloo, morning :) | 13:12 |
| dtantsur | rloo, morning | 13:12 |
| rloo | yuriyz: wrt https://review.openstack.org/123150, why aren't port/node destroys related to that bug? | 13:12 |
| rloo | hiya lucasagomes, dtantsur | 13:12 |
| *** vdrok has joined #openstack-ironic | 13:16 | |
| jroll | dtantsur: oh yeah, go for it :) | 13:19 |
| yuriyz | rloo, DB API can raise NotFound for port/node, but REST API does get_by_uuid() before (uuid will be in error message), only ChassisNotEmpty exception related to this bug ("id" in the message) | 13:20 |
| rloo | yuriyz: hmm. ok, I trust that you're right since I'm not familiar with that code. I don't like that the code doesn't seem consistent now, but I guess that's another story. Thx. | 13:22 |
| *** nosnos has quit IRC | 13:27 | |
| *** nosnos has joined #openstack-ironic | 13:27 | |
| *** dprince has joined #openstack-ironic | 13:31 | |
| *** nosnos has quit IRC | 13:32 | |
| yuriyz | rloo, now I think we should change obj.destroy() for port/node also, because this bug will be in delete() race case. Thanks. | 13:35 |
| rloo | yuriyz: thx! ;) | 13:38 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Separate the agent driver config from the base localrc config https://review.openstack.org/123733 | 13:46 |
| lucasagomes | jroll, ^ to facilitate the c&p | 13:47 |
| openstackgerrit | Yuriy Zveryanskyy proposed a change to openstack/ironic: Do not return 'id' in REST API error messages https://review.openstack.org/123150 | 13:47 |
| * lucasagomes built an env to test the pxe hash ring but ended with agent_ssh enabled by default | 13:47 | |
| jroll | heh, oops | 13:48 |
| jroll | +2'd | 13:48 |
| *** r-daneel_ has joined #openstack-ironic | 13:59 | |
| *** bluex-pl has joined #openstack-ironic | 14:07 | |
| NobodyCam | good morning Ironic | 14:21 |
| lucasagomes | NobodyCam, morning | 14:21 |
| NobodyCam | morning lucasagomes | 14:21 |
| NobodyCam | morning rloo :) | 14:21 |
| rloo | bonjour NobodyCam | 14:22 |
| NobodyCam | yuriyz: morning | 14:22 |
| NobodyCam | :) | 14:22 |
| yuriyz | morning NobodyCam | 14:22 |
| NobodyCam | yuriyz: I had a question about https://review.openstack.org/#/c/103105 | 14:23 |
| NobodyCam | are you still working on that one? | 14:23 |
| NobodyCam | :) | 14:23 |
| Shrews | morning Mr. NobodyCam | 14:24 |
| NobodyCam | morning Shrews :) | 14:24 |
| yuriyz | NobodyCam, yes I should add some doc and not proper tested yet | 14:24 |
| NobodyCam | I also added a couple of comments | 14:24 |
| yuriyz | NobodyCam, thanks | 14:25 |
| jroll | morning NobodyCam and yuriyz :) | 14:25 |
| NobodyCam | I don't get lines 12 thru 17 of https://review.openstack.org/#/c/103105/5/elements/ironic-agent/post-install.d/80-ironic-agent | 14:25 |
| NobodyCam | morning jroll | 14:25 |
| NobodyCam | are those lines needed | 14:25 |
| jroll | yuriyz: does that IPA DIB thing start the agent properly and everything? eager to try it | 14:26 |
| yuriyz | jroll, deploy not tested yet | 14:27 |
| jroll | ok | 14:28 |
| jroll | I have another thing to do but I think I'll try that this morning :) | 14:28 |
| yuriyz | jroll, thanks | 14:28 |
| NobodyCam | awesome jroll :) | 14:29 |
| * jroll starts building it now | 14:30 | |
| jroll | one thing that would be nice is to be able to build this from source | 14:30 |
| jroll | dunno how feasible that is | 14:30 |
| *** bluex-pl has quit IRC | 14:42 | |
| *** bluex-pl has joined #openstack-ironic | 14:43 | |
| *** pcrews has joined #openstack-ironic | 14:47 | |
| NobodyCam | brb | 14:52 |
| *** Poornima has joined #openstack-ironic | 14:57 | |
| *** ifarkas has quit IRC | 14:59 | |
| NobodyCam | rain rain go away gome again some other day! | 15:02 |
| NobodyCam | :-p | 15:02 |
| *** ramineni has joined #openstack-ironic | 15:10 | |
| NobodyCam | no code for https://bugs.launchpad.net/ironic/+bug/1365334 | 15:10 |
| NobodyCam | :( | 15:10 |
| lucasagomes | hmm there were some | 15:13 |
| lucasagomes | it's in the queue afair | 15:14 |
| NobodyCam | I will dig in in a minute | 15:14 |
| Shrews | it was something from lifeless, iirc | 15:15 |
| Shrews | https://review.openstack.org/118932 | 15:15 |
| NobodyCam | dtantsur: you Abandoned https://review.openstack.org/#/c/92115 and https://review.openstack.org/#/c/93748?? | 15:15 |
| NobodyCam | Shrews: TY, odd it didn't get added to the bug | 15:16 |
| Shrews | NobodyCam: bug came after the fix | 15:17 |
| Shrews | but, hrm... there is a closes-bug tag on it. | 15:18 |
| Shrews | weird | 15:18 |
| * Shrews blames LP | 15:18 | |
| rloo | Shrews: I just gave feedback on 123518, but I decided not to look at it all, cuz I should probably look at code changes for RC1. | 15:18 |
| *** jistr has quit IRC | 15:21 | |
| dtantsur | NobodyCam, yeah, IIRC I got message from devananda that it's not the right way to do it | 15:26 |
| NobodyCam | I was reading his comments | 15:28 |
| *** rushiagr is now known as rushiagr_away | 15:29 | |
| *** loki184 has joined #openstack-ironic | 15:30 | |
| Shrews | rloo: so, i think you are going to force me to outline how i think doc reviews should be handled :) | 15:30 |
| rloo | Shrews: sorry. How should they be handled? | 15:31 |
| Shrews | rloo: tl;dr... IMO, docs should be reviewed more for correctness vs. formatting/completeness. Especially when it's for documentation that is non-existent. Something is better than nothing. | 15:32 |
| *** ramineni has quit IRC | 15:32 | |
| rloo | Shrews: so you're saying, first pass for new docn is for correctness of info. then let later revisions fix formatting/completeness? | 15:33 |
| Shrews | pretty much, yeah. documentation should be a collaborative effort since no single person wants to do it. | 15:33 |
| rloo | Shrews: I'm fine with that if others are too. Although I'm not sure I'm really fine with that, but if others are fine with that... | 15:33 |
| NobodyCam | oh nice: https://review.openstack.org/#/c/93748 | 15:33 |
| *** lazy_prince is now known as killer_prince | 15:34 | |
| NobodyCam | Shrews: +1 to that.. Something is better then nothing! | 15:35 |
| Shrews | rloo: for example, you prefer an alternate organization of the info, but i like it the way i presented it. should we stall the docs until we come to a consensus, or go ahead and put it in so that at least there is SOMETHING there? | 15:35 |
| rloo | Shrews: but that's the discussion we have with all reviews, if there is diff of opinion. Isn't it better to work it out before something goes in, than try to fix/modify afterwards? Even afterwards, I hope the discussion still happens. | 15:36 |
| rloo | Shrews: it just seems more efficient to me to do it 'as best' the first time, than to then have to deal with subsequent patches. | 15:36 |
| Shrews | rloo: i think code is totally different. code MUST work and MUST be as close to perfect we can get it. refactoring is hard | 15:36 |
| *** rushiagr_away is now known as rushiagr | 15:37 | |
| Shrews | code can break us, docs can't :) | 15:37 |
| * NobodyCam is not sure he agrees with Shrews on that | 15:37 | |
| rloo | Shrews: ohhh kay. But as someone that reads doc, I like to be able to read/understand what I'm reading. | 15:38 |
| rloo | Shrews: Anyway, I know I can be picky. So I'm willing to not be picky if that's the consensus. | 15:38 |
| Shrews | rloo: i actually agree with you that *ideally* it would be perfect the first time. the problem is we have no standard of "perfect" for docs. | 15:38 |
| Shrews | rloo: i would never suggest accepting docs that are not understandable | 15:39 |
| rloo | Shrews: yes, there is no standard. So I try to review the doc as if I were the reader w/o much knowledge. | 15:39 |
| rloo | Shrews: so my comments were based on understandability. | 15:40 |
| rloo | Shrews: "my" understandability. Or how it can be done better to make it easier for me to understand. | 15:40 |
| rloo | Shrews: clearly everyone has diff opinions on what goes in doc. you can see that by the comments you get. | 15:40 |
| Shrews | rloo: indeed. :) so, i'm not clear what it is about "options" that's confusing you? | 15:42 |
| rloo | Shrews: So I guess I wasn't clear in my comments. The first section was title 'Options', and it referred to the optional arguments, the --* stuff. | 15:43 |
| rloo | Shrews: the second section starts off with a sentence that mentions 'options' and to do an ironic-dbsync -h to see them all. But you had already mentioned that in the first section. | 15:44 |
| rloo | Shrews: So my thinking is maybe you meant something else in the second section wrt the term 'options'. | 15:44 |
| rloo | Shrews: and since the second section was titled 'usage', I sort of assumed it'd show the usage stuff. | 15:44 |
| jroll | yuriyz: update, I can't get this thing to boot :/ | 15:45 |
| rloo | Shrews: like when you look at a man page. | 15:45 |
| Shrews | rloo: no, that was just a repeat. showing how to "use" help :) | 15:45 |
| *** ndipanov is now known as ndipanov_gone | 15:45 | |
| Shrews | rloo: i can remove that though, since it's a repeat | 15:46 |
| rloo | Shrews: oh. Well then, yeah, remove it if you were only talking about the optional arguments. | 15:46 |
| jroll | yuriyz: but not entirely sure why... going to look a bit | 15:46 |
| rloo | Shrews: but the help shows the usage/args/subcommands, so more completely, you can say 'use help to see the full something of the command' | 15:47 |
| *** viktors is now known as viktors|afk | 15:47 | |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic: Add documentation for ironic-dbsync command https://review.openstack.org/123518 | 15:51 |
| Shrews | rloo: not sure i understood that last bit, but ^^^^ | 15:51 |
| rloo | Shrews: I want to focus on the hash ring stuff today (was getting distracted with other stuff). So I'll take a look at this later. docn can be updated after rc1 cuz they just show up when they're merged. | 15:53 |
| Shrews | rloo: certainly! | 15:53 |
| loki184 | hi all had a question using host aggregates creation with ironic enabled multinode devstack setup | 15:54 |
| rloo | Shrews: one question I have for you. If the doc for the commands doesn't show everything that the command can do, why/what subset is being documented, and it should be clear that the doc isn't complete. | 15:54 |
| NobodyCam | Shrews: may I ask why Revision, stamp and version do not have command examples as the others do? | 15:54 |
| loki184 | can someone please help me with grouping normal vms vs ironic baremetal vms using host aggregates and scheduler filter | 15:54 |
| openstackgerrit | A change was merged to openstack/ironic: Separate the agent driver config from the base localrc config https://review.openstack.org/123733 | 15:54 |
| jroll | hmm | 15:54 |
| * jroll wonders who knows about host aggregates and such | 15:55 | |
| Shrews | rloo: i think i did document all of the subcommand options. not sure what you're asking | 15:55 |
| rloo | Shrews: your comment 'I intentionally did not add that because then we'd force documentation update for any change in the options | 15:55 |
| jroll | loki184: I'd love to help, but I don't use host aggregates, my team runs ironic in a separate cell | 15:55 |
| Shrews | NobodyCam: b/c i didn't think all of them (like version) needed an example. and i plain just don't know why or how i'd use stamp. :) | 15:55 |
| NobodyCam | loki184: I have not ever setup such a configuration, but would welcome a doc on how to do it | 15:55 |
| NobodyCam | lol | 15:56 |
| NobodyCam | Shrews: ack | 15:56 |
| jroll | loki184: might be best to mail the list so that nova folks can chime in (they might know more) | 15:56 |
| loki184 | jroll: i have setup with two computehost service running want to provision vms on one of the host as per user choice. i will check with nova folks thanks! | 15:57 |
| loki184 | NobodyCam: Will document it :) | 15:57 |
| NobodyCam | loki184: awesome Thank you :) | 15:57 |
| Shrews | rloo: the subset i mentioned are what i thought would be the most used options (the subcommands list the full options for that command) and that a full list is available with -h | 15:58 |
| *** killer_prince is now known as lazy_prince | 15:59 | |
| *** Poornima has quit IRC | 16:00 | |
| rloo | Shrews: yes. hmm. I guess I really like to see the usage string from the command. | 16:00 |
| devananda | morning, all | 16:01 |
| jroll | heya devananda :) | 16:01 |
| NobodyCam | good morning devananda | 16:01 |
| rloo | morning devananda | 16:02 |
| devananda | zigo: my testing yesterday indicates that the current requirements (0.2.1) are sufficient for the nova driver, but the latest release (0.3.0) does have several improvements | 16:02 |
| *** k4n0 has quit IRC | 16:03 | |
| lucasagomes | devananda, morning | 16:03 |
| zigo | devananda: I'll try to find the time to update the package to 0.3.0 then. | 16:03 |
| lucasagomes | devananda, do we have a consensus if we are going to fix the take over in J? I started looking into that | 16:04 |
| *** saurabhs has joined #openstack-ironic | 16:04 | |
| lucasagomes | devananda, I'd like to validate some assumptions, for e.g: Is it fine to only call a take over for nodes that are *deployed* | 16:05 |
| lucasagomes | > | 16:05 |
| lucasagomes | ?* | 16:05 |
| devananda | zigo: thanks much | 16:05 |
| devananda | lucasagomes: correct | 16:05 |
| lucasagomes | it seems so... because I want to make it a light operation | 16:06 |
| lucasagomes | right | 16:06 |
| devananda | also, https://review.openstack.org/#/c/122819/5/ironic/drivers/modules/ilo/deploy.py seems wierd | 16:06 |
| lucasagomes | I want to create a bound conductor->node as part of the deployment of the node, and remove it when we undeploy | 16:06 |
| devananda | why is ilo management using ipmitool? | 16:06 |
| *** rameshg87 has joined #openstack-ironic | 16:06 | |
| lucasagomes | devananda, to be able to set the boot device? | 16:06 |
| lucasagomes | and get sensor data | 16:07 |
| lucasagomes | etc | 16:07 |
| devananda | right - why isn't it doing that with *ilo* ? | 16:07 |
| * lucasagomes thinks (that sounds a better idea really) | 16:08 | |
| devananda | this means a user of the ilo driver needs to input both ilo and ipmi credentials | 16:08 |
| devananda | and make sure both proliantutils and ipmitool are present on teh system | 16:09 |
| lucasagomes | yeah, they were using the ipmitool.Management directly before | 16:09 |
| lucasagomes | and this is something we talked when the ilo driver was being proposed afair | 16:09 |
| lucasagomes | and they were about it, I think that's was when the "credentials should be separated" topic came from | 16:10 |
| devananda | the commit message doesn't match the implementation at all | 16:10 |
| devananda | This fix proposes to change the management interface for pxe_ilo driver | 16:10 |
| devananda | to call ilo set_boot_device using proliantutils ilo_client functions | 16:10 |
| devananda | instead of IPMI call | 16:10 |
| devananda | ... but it's not using ilo_client | 16:10 |
| devananda | wait, is it? | 16:11 |
| lucasagomes | it's | 16:11 |
| lucasagomes | ilo_common.set_boot_device(task.node, boot_device, persistent) | 16:11 |
| devananda | heh, it is | 16:11 |
| devananda | yea | 16:11 |
| lucasagomes | before it was using ipmitool | 16:11 |
| devananda | it's still subclassing IPMIManagement tho | 16:11 |
| *** bluex-pl has quit IRC | 16:12 | |
| devananda | which includes validate() | 16:12 |
| lucasagomes | also https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ilo/common.py#L262 | 16:13 |
| lucasagomes | devananda, yeah, that's not ideal | 16:13 |
| lucasagomes | but not worse than it was before too | 16:13 |
| *** rameshg87 has left #openstack-ironic | 16:14 | |
| devananda | anyone know if Syed is on IRC? | 16:14 |
| devananda | I dont think I know his/her handle | 16:14 |
| lucasagomes | ramineni (I think) | 16:15 |
| lucasagomes | he was the one pinging us to take a look at those patches | 16:15 |
| devananda | wanyen: ping | 16:15 |
| * devananda files a new bug | 16:16 | |
| devananda | dtantsur: https://bugs.launchpad.net/ironic/+bug/1367182 isn't invalid just because we're in feature freeze | 16:17 |
| devananda | dtantsur: it's still a bug, and actually, a fairly bad one IMHO | 16:19 |
| dtantsur | devananda, I put it as "invalid" because we're decided some time ago to have a good talk about credentials unification | 16:21 |
| devananda | dtantsur: sure - but "we'll talk about how to fix this later" doesn' tmake the current problem invalid | 16:21 |
| *** christopheraedo has quit IRC | 16:21 | |
| devananda | that the iLO driver also requires IPMITool's credentials is a bug, IMO | 16:22 |
| devananda | they've proposed a solution in this bug report so that, from a user POV, it won't | 16:22 |
| devananda | and so that, if we do consolidate credentials in teh future, this will also make that easier and not break existing users | 16:22 |
| dtantsur | devananda, yeah, it's a complex thing. I just didn't want the reporter to end up with proposing patch and hear "no, please not now and not that way" | 16:23 |
| dtantsur | devananda, then we need to change both credentials to just 'username' and 'password' | 16:23 |
| NobodyCam | lucasagomes: your looking at the has ring stuff? | 16:23 |
| lucasagomes | NobodyCam, yup | 16:23 |
| dtantsur | devananda, it would be the start of consolidation, but without a spec and an agreement | 16:23 |
| lucasagomes | NobodyCam, well, I'm first looking at how I'll bound nodes to conductors, and get a delta of what changed when it comes to rebalance it and trigger take_over on the nodes that have been affected | 16:24 |
| NobodyCam | I addeda couple of old patches to the etherpad | 16:24 |
| devananda | dtantsur: sure, that's a valid point. | 16:24 |
| lucasagomes | NobodyCam, greg ones? will take a look | 16:24 |
| lucasagomes | thanks! | 16:24 |
| dtantsur | devananda, honestly I never know what to do with this kind of bugs :( and I'm seeking your guidance | 16:24 |
| lucasagomes | NobodyCam, ohh no no I'm not looking into it | 16:25 |
| dtantsur | devananda, how to say "right, but please don't take actions w/o spec and an agreement" | 16:25 |
| NobodyCam | ahh ok :) | 16:25 |
| lucasagomes | NobodyCam, I'm looking more into https://bugs.launchpad.net/bugs/1355510 | 16:25 |
| devananda | dtantsur: in those cases, the mailing list is probably the best answer | 16:25 |
| lucasagomes | NobodyCam, but that hash ring problem affects it because if it have to move a lot of nodes around the take over is more costy | 16:26 |
| devananda | dtantsur: since that hits a wider audience and gives people in different timezones the opportunity to participate | 16:26 |
| dtantsur | well yeah, maybe | 16:26 |
| devananda | dtantsur: I'm at fault for not using the ML enough | 16:26 |
| * lucasagomes also needs to use ML more/better | 16:26 | |
| dtantsur | devananda, right, we didn't have any serious holy war on ML! Something to fix in K cycle :) | 16:27 |
| devananda | dtantsur: haha! yes :) | 16:27 |
| *** christopheraedo has joined #openstack-ironic | 16:28 | |
| jroll | hahaha | 16:32 |
| *** r-daneel_ has quit IRC | 16:33 | |
| *** openstackgerrit has quit IRC | 16:35 | |
| *** romcheg has quit IRC | 16:35 | |
| devananda | lucasagomes: i'm pushing other distractions aside for a bit, if you're still around and want to chat about the hash ring | 16:36 |
| lucasagomes | devananda, yeah I'm here, I'm investigating it yet... trying to make some assumptions | 16:37 |
| lucasagomes | devananda, but sometimes I think that another way to solve it for the pxe driver specifically | 16:37 |
| lucasagomes | would be to generate the configs on the fly | 16:37 |
| devananda | ? | 16:37 |
| lucasagomes | (it's possible with ipxe) | 16:37 |
| devananda | which configs | 16:37 |
| lucasagomes | pxe config | 16:37 |
| JayF | ah, like you'd have the pxe config come from http://ironic-api/some/path/ipxe | 16:37 |
| lucasagomes | and make the pxe driver stateless | 16:37 |
| lucasagomes | JayF, yes | 16:38 |
| lucasagomes | I'm think about the short term (tho that's a good solution) | 16:38 |
| lucasagomes | so we can be able to fix it in J rc2 | 16:38 |
| devananda | lucasagomes: stateless? go on | 16:39 |
| lucasagomes | devananda, yeah... the boot process of ipxe is scripted, so we can make it do a call to the ironic api when it's booting passing the mac address | 16:40 |
| lucasagomes | the ironic get the mac and find the node | 16:40 |
| lucasagomes | look at the provision_state of the node | 16:40 |
| lucasagomes | if it's DEPLOYWAIT for e.g the deploy ramdisk should be booted | 16:40 |
| lucasagomes | so we tell the ipxe script to boot it | 16:41 |
| lucasagomes | if it's ACTIVE we tell the ipxe script to boot the image ramdisk | 16:41 |
| devananda | that'd remove the dependency on neutron | 16:41 |
| devananda | since we wouldn't need the DHCP service to respond with the IP address of a specific conductor | 16:41 |
| lucasagomes | so we don't keep config files locally | 16:41 |
| lucasagomes | and have to regenetated it etc... | 16:41 |
| JayF | If this were implemented | 16:42 |
| lucasagomes | devananda, thing is... we still need to bootstrap from normal pxe to ipxe in some cases | 16:42 |
| JayF | I'd suspect we'd move our production over to use these, because it's awesome | 16:42 |
| lucasagomes | so neutron is needed for that bit | 16:42 |
| JayF | lucasagomes: you can do that with a simple static dhcp config | 16:42 |
| jroll | JayF++ | 16:42 |
| JayF | lucasagomes: I love this because you're detaching the dhcp from the dynamic stuff | 16:42 |
| devananda | we would need ipxe to fetch the boot config from the Ironic API service, which, of course, would be load balanced, and could point the rest of the iPXE script to any conductor at that time | 16:43 |
| JayF | lucasagomes: I want Ironic to be able to tell the node things when it boots without having to own the dhcp server...and this would do it | 16:43 |
| lucasagomes | JayF, yup! | 16:43 |
| jroll | lucasagomes, devananda, this will also help support rescue mode :) | 16:43 |
| devananda | lucasagomes: the dhcp service would need to respond to teh PXE boot request with the iPXE kernel location, yes, but taht's static -- not dynamic per MAC address | 16:43 |
| jroll | (I think) | 16:43 |
| lucasagomes | devananda, yeah, that's static | 16:44 |
| devananda | lucasagomes: this sounds great. it's a massive change this late in the cycle, tho :( | 16:44 |
| lucasagomes | but this does not solve the *general* take over problem | 16:44 |
| lucasagomes | which is intent to be generic | 16:44 |
| lucasagomes | take_over still not being called when the ring is rebalanced for the nodes affected | 16:44 |
| lucasagomes | devananda, yeah :( | 16:44 |
| JayF | lucasagomes: something like this could even be used in IPA itself... we've talked about (me and jroll) making the ramdisk image mostly static... where for CoreOS ramdisk users, they'd just run a "stock" coreos pxe image, and be fed a cloud-config.yml via kernel command line on boot that tells it to download and extract IPA along with all the needed deps | 16:44 |
| lucasagomes | devananda, JayF should I write a spec about it for K | 16:45 |
| lucasagomes | if u guys are interested I think it's a good addition I can work on it | 16:45 |
| JayF | Very very yes | 16:45 |
| devananda | lucasagomes: +100 | 16:45 |
| JayF | please make it generic enough that ipa can use it as well | 16:45 |
| *** jistr has joined #openstack-ironic | 16:45 | |
| lucasagomes | ack I will ;) /me writes it down | 16:45 |
| JayF | and maybe even a mechanism for dynamically generating OTHER configs would be useful too | 16:45 |
| lucasagomes | JayF, oh yeah, sure I will try to be as generic as possible | 16:45 |
| devananda | I'm wondering if there's ever another need for takeover to rebuild local state | 16:46 |
| lucasagomes | devananda, maybe for other drivers... I think take cover is cool | 16:46 |
| devananda | if configs are generated on the fly by the API service, rather than in-advance | 16:46 |
| lucasagomes | but I can't see how to make it not being so costy | 16:46 |
| JayF | the only thing is | 16:47 |
| lucasagomes | without having to loop around loads of nodes | 16:47 |
| JayF | for non-ipxe users | 16:47 |
| JayF | you can't generate the config on the fly | 16:47 |
| lucasagomes | yeah | 16:47 |
| devananda | right | 16:47 |
| JayF | because it has to be tftp'able | 16:47 |
| lucasagomes | that's also ^ | 16:47 |
| JayF | That being said who's in for deprecating PXELINUX support | 16:47 |
| lucasagomes | I'd like also to make ipxe default if possible | 16:47 |
| devananda | yea, for hardwar that doesn't support ipxe ... | 16:47 |
| JayF | o/ | 16:47 |
| devananda | does that exist? :) | 16:47 |
| lucasagomes | I don't see reason why use PXE | 16:47 |
| lucasagomes | JayF, +1 | 16:47 |
| JayF | devananda: you can bootstrap any UNDI-supporting hardware into ipxe | 16:47 |
| lucasagomes | devananda, if it doesn't support iPXE it can bootstrap the image | 16:47 |
| lucasagomes | from the dhcp server | 16:47 |
| devananda | right | 16:48 |
| lucasagomes | yeah like JayF said, but with more specific terms | 16:48 |
| JayF | iPXE also has some more hardware-specific ROMs | 16:48 |
| NobodyCam | didn't we have a spec fir just that? | 16:48 |
| JayF | but those are generally used for flashing on the chip | 16:48 |
| JayF | I don't forsee a case where a standard pxelinux.0 rom would work that an ipxe undi rom wouldn't | 16:49 |
| lucasagomes | NobodyCam, deprecating PXE? not that I know | 16:49 |
| JayF | NobodyCam: lucasagomes added support for ipxe, but optional and I don't even think default support | 16:49 |
| devananda | AIUI, the iLO driver stores its images in Swift, so also doesn't need local state | 16:49 |
| lucasagomes | devananda, yeah, ipxe can boot from http | 16:50 |
| NobodyCam | ack that was it | 16:50 |
| lucasagomes | so we could just pass the url | 16:50 |
| lucasagomes | temp url* | 16:50 |
| JayF | we boot from http using ipxe all the time :) | 16:50 |
| lucasagomes | and boot from that too, and avoid caching locally | 16:50 |
| *** romcheg has joined #openstack-ironic | 16:50 | |
| lucasagomes | the downside I see is that, we would depend on a http server to be up and running (just like we now depend on a tftp server) | 16:50 |
| devananda | lucasagomes: that's an improvement, IMO | 16:51 |
| JayF | Ironic /is/ an http server | 16:51 |
| devananda | http is much more reliable | 16:51 |
| NobodyCam | jbjohnso__: was in support of aways chain loading ipxe thru pxe | 16:51 |
| NobodyCam | if I recall correctly | 16:51 |
| JayF | lucasagomes: I don't think you should depend on ANOTHER http server | 16:51 |
| lucasagomes | JayF, oh yeah... that's also, because now I put the images on a HTTP server | 16:51 |
| devananda | JayF: though fwiw, i don't want our API server to take on image streaming ... | 16:51 |
| lucasagomes | but with the use of swift | 16:51 |
| JayF | lucasagomes: ironic already has an http api; why shouldn't these just be capability urls? | 16:51 |
| lucasagomes | we don't need another anymore | 16:51 |
| lucasagomes | JayF, +1! u right | 16:51 |
| JayF | devananda: I agree, but this is a simple config file in this particular case :) If people need images they should talk to glance or swift | 16:52 |
| devananda | JayF: because then it also has to take on receiving and storing those images | 16:52 |
| devananda | JayF: right - swift | 16:52 |
| JayF | what images are you talking about? | 16:52 |
| *** pensu has quit IRC | 16:52 | |
| lucasagomes | devananda, JayF all that together is a lot of work... I see at least 2 bps here... 1 to make iPXE default. Then another implementing the generation of the config files and using swift etc | 16:52 |
| JayF | we're talking about a text file that'd likely be five lines | 16:52 |
| devananda | JayF: don't you guys have a swift cluster or something already? :p | 16:52 |
| JayF | and generated on the fly | 16:53 |
| devananda | JayF: oh, no -- THAT config file absolutely should be served by ironic-api | 16:53 |
| JayF | the ipxe image has to be served over tftp, if that's what you mean | 16:53 |
| devananda | JayF: I meant the actual image endpoint, which should also be fetched over http instead of tftp | 16:53 |
| JayF | because it has to be bootstrapped in using the normal pxe process unless you flash your own nic | 16:53 |
| JayF | devananda: I missed something, because I'm not sure where images came into play in the conversation? | 16:54 |
| lucasagomes | yeah the ipxe image should be served by tftp | 16:54 |
| devananda | 16:49:26 < devananda> AIUI, the iLO driver stores its images in Swift, so also doesn't need local state | 16:55 |
| devananda | ... | 16:55 |
| devananda | 16:50:53 < lucasagomes> the downside I see is that, we would depend on a http server to be up and running (just like we now depend on a tftp server) | 16:55 |
| lucasagomes | tho, if all ur hardware already support iPXE we can make it optional | 16:55 |
| devananda | JayF: that section of conversation | 16:55 |
| JayF | Ah, okay, gotcha. Absolutely those belong and go into swift :) | 16:55 |
| devananda | :) | 16:55 |
| devananda | so, to summarize | 16:55 |
| devananda | - tftp still needs to serve the iPXE ROM (since we can't require all operators to flash that into firmware) | 16:56 |
| devananda | - deploy and user kernel and ramdisk images can be fetched over HTTP (from swift) rather than over TFTP | 16:56 |
| *** harlowja_away is now known as harlowja | 16:56 | |
| *** derekh has quit IRC | 16:56 | |
| devananda | - iPXE config should be served by ironic-api, making it dynamically-generatable, and removing the need to maintain it as local-state on a given conductor | 16:56 |
| lucasagomes | yeah | 16:57 |
| JayF | bingo, this is great | 16:57 |
| lucasagomes | and we have to think about what to do with take_over() :) | 16:57 |
| devananda | - then, outside of the times where a conductor has a Task *lock* on a node, there is no dependency between a given node and a given conductor | 16:57 |
| lucasagomes | yeah, for the drivers we have now in tree it sounds correct | 16:58 |
| lucasagomes | I still see take_over() as a valid operation tho | 16:58 |
| lucasagomes | I don't think we should dump it... | 16:58 |
| devananda | (though, if conductor 1 starts a deploy, and conductor 2 continues it, c2 may need to fetch the user image from glance at that point in time) | 16:58 |
| * lucasagomes just can't think about how to make it a light operation | 16:59 | |
| lucasagomes | devananda, yeah to write it to the disk | 16:59 |
| devananda | I think the above could become *the* PXE implementation, and the same processes should work for IPA | 17:00 |
| devananda | along with the refactoring of the deploy vs boot mechanism | 17:00 |
| lucasagomes | yeah it would be pretty good indeed, I will start a spec about it | 17:00 |
| devananda | we end up with a PXE boot driver that is shared by both | 17:00 |
| devananda | and an iSCSI deploy + an Agent deploy | 17:00 |
| devananda | JayF: yes? | 17:00 |
| lucasagomes | yeah, good point | 17:00 |
| lucasagomes | that separation of the boot interface would be great to happen as soon as possible when K opens | 17:01 |
| JayF | one note: I'd call it the iPXE driver, simply because it's more accurate and avoids overloading "PXE" any more in Ironic :) | 17:01 |
| devananda | JayF: fair point | 17:01 |
| JayF | telling people how the agent pxes is sometimes like doing a whos on first skit | 17:01 |
| devananda | heh | 17:01 |
| lucasagomes | devananda, JayF you guys want a session about it in paris? | 17:02 |
| devananda | lucasagomes: yes | 17:02 |
| lucasagomes | or we are pretty much in agreement and it's not needed? | 17:02 |
| JayF | we are three :) | 17:02 |
| lucasagomes | heh | 17:02 |
| devananda | even if we are in agreement, it needs a session | 17:02 |
| JayF | I can pretty much say jroll would be in too because he's been wanting this for a while :P | 17:02 |
| lucasagomes | ack ok | 17:02 |
| lucasagomes | I will propose that then | 17:02 |
| * lucasagomes writes everything down to not forget | 17:02 | |
| devananda | it's a fairly major change in the default driver - even if everyone agrees that it's good, we will need community awareness | 17:03 |
| jroll | ? | 17:03 |
| * jroll reads | 17:03 | |
| lucasagomes | devananda, indeed | 17:03 |
| jroll | JayF++ do want | 17:03 |
| *** rwsu has quit IRC | 17:03 | |
| lucasagomes | devananda, btw, maybe we even should start a new driver as JayF said | 17:03 |
| lucasagomes | iPXE driver | 17:03 |
| lucasagomes | instead of making it as we do today, PXE and IPXE (changing a config) | 17:03 |
| devananda | considering our need to support existing users, give them time to migrate, etc | 17:03 |
| devananda | yep | 17:04 |
| devananda | we can't delete or suddenly change teh functionality of an existing driver | 17:04 |
| lucasagomes | yeah | 17:04 |
| lucasagomes | right | 17:04 |
| lucasagomes | devananda, so, for J | 17:04 |
| lucasagomes | what we do with the take_over() | 17:04 |
| lucasagomes | ? | 17:05 |
| *** rwsu has joined #openstack-ironic | 17:06 | |
| lucasagomes | I mean, anyone... ideas? | 17:06 |
| *** rushiagr is now known as rushiagr_away | 17:08 | |
| *** todd_dsm has joined #openstack-ironic | 17:10 | |
| devananda | hackish idea - simple periodic task that refreshes the local ring | 17:11 |
| devananda | well, not so hackish, i mean, i think that's what i wanted to do originally :) | 17:12 |
| lucasagomes | right, yeah minimize the problem | 17:13 |
| lucasagomes | maybe for J we should really work on the rebalancing problem of the hash ring | 17:13 |
| lucasagomes | + that periodic task | 17:13 |
| lucasagomes | it's not perfect but it's seems reasonable for the time we have | 17:14 |
| NobodyCam | brb .... quick walkies | 17:14 |
| devananda | the redistribution is just silly, and probably my fault | 17:14 |
| devananda | i only tested it with small rings, at which scale that doesn't really show up | 17:14 |
| devananda | i didn't test the distribution of a ring with 10k conductors :( | 17:14 |
| lucasagomes | devananda, well, we reviewed that, so it's everyone's fault :) | 17:14 |
| lucasagomes | alright, I will think about it tomorrow a bit more | 17:16 |
| lucasagomes | it's late here and I'm hungry | 17:16 |
| lucasagomes | have a good night everyone | 17:17 |
| lucasagomes | devananda, last thing... we don't know how many slots we have for the design summit yet do we? | 17:17 |
| romcheg | gnight lucasagomes! | 17:17 |
| romcheg | And good evening too :) | 17:17 |
| lucasagomes | romcheg, night :) | 17:17 |
| romcheg | Hi everyone | 17:17 |
| devananda | lucasagomes: nope | 17:17 |
| devananda | romcheg: hi! | 17:18 |
| devananda | lucasagomes: g'night! I'll update the etherpad as I go. | 17:18 |
| devananda | oh - I'm giogn to be offline most of tomorrow | 17:18 |
| lucasagomes | alright thanks! | 17:18 |
| devananda | pulled into a meeting in sunnyvale... | 17:18 |
| lucasagomes | devananda, oh ok... I will try to start a spec about the iPXE stuff we talked, and a presentation for the design summit | 17:18 |
| romcheg | Hope after I fix all red tape here I'll be able to spend some time with you | 17:18 |
| lucasagomes | devananda, alright :) enjoy the meeting there | 17:18 |
| lucasagomes | devananda, you should go to that bar with the pool table, oasis! | 17:19 |
| lucasagomes | that was a cool bar | 17:19 |
| *** lucasagomes is now known as lucas-dinner | 17:20 | |
| NobodyCam | night lucas-dinner | 17:22 |
| NobodyCam | hey romcheg :) | 17:22 |
| romcheg | Hi devananda, NobodyCam! | 17:23 |
| NobodyCam | hey romcheg how's the NEW pad? | 17:23 |
| romcheg | NobodyCam: Well, not bad at all except I'm not in the upstream now | 17:24 |
| NobodyCam | :/ | 17:25 |
| *** igordcard has joined #openstack-ironic | 17:25 | |
| *** openstackgerrit has joined #openstack-ironic | 17:25 | |
| romcheg | But here I'll have more free time so I plan to spend some time with you as soon as I manage to deal with all red tape for me and my wife | 17:25 |
| NobodyCam | :) | 17:26 |
| *** harlowja has quit IRC | 17:27 | |
| romcheg | Now I barely manage to do my paid job because of that :) | 17:27 |
| *** harlowja has joined #openstack-ironic | 17:27 | |
| NobodyCam | :( | 17:27 |
| *** dtantsur is now known as dtantsur|afk | 17:28 | |
| NobodyCam | devananda: anything need covering when you in sunnyvale? | 17:28 |
| romcheg | medical exams, apartment, n+1 different registrations, taxes — that's what every immigrant has to do during their first days | 17:29 |
| NobodyCam | romcheg: gah, | 17:29 |
| dtantsur|afk | romcheg, oh yeah, it's all fun | 17:29 |
| romcheg | dtantsur|afk: you know what I mean :) | 17:30 |
| dtantsur|afk | though I still have to deal with taxes next year | 17:30 |
| dtantsur|afk | romcheg, do you at least know the language people around use? | 17:30 |
| romcheg | dtantsur|afk: I undertand Polish but don't speak it. So I use Ukrainian + improvisation to convert it to Polish. It's not very hard :) | 17:31 |
| dtantsur|afk | romcheg, heh better than for me, I understood nearly no Czech back then | 17:32 |
| *** dtantsur|afk is now known as dtantsur|nearly- | 17:32 | |
| NobodyCam | lol that sounds like a utube video in the making | 17:32 |
| *** dtantsur|nearly- is now known as dtantsur|nearlya | 17:32 | |
| Shrews | devananda: so, RC's should start being proposed beginning tomorrow? | 17:32 |
| Shrews | https://wiki.openstack.org/wiki/Juno_Release_Schedule | 17:32 |
| *** dtantsur|nearlya is now known as dtantsur|afk | 17:32 | |
| dtantsur|afk | ok afk for real, see you | 17:33 |
| rloo | ciao dtantsur|afk | 17:33 |
| romcheg | dtantsur|afk: bye | 17:33 |
| JayF | Shrews: aiui, from what's been said in channel, we're not supposed to cut an RC until any known bugs we'd be comfortable releasing with are gone | 17:34 |
| JayF | Shrews: so I think we're waiting for a hash ring fix before cutting rc1 but imbw | 17:34 |
| *** marcoemorais has joined #openstack-ironic | 17:35 | |
| Shrews | JayF: devananda: i ask b/c one of our targeted bugs, 1271317, is assigned to mrda and he's still on his extended vacation. just wondering about that one | 17:35 |
| NobodyCam | have a good night dtantsur|afk | 17:35 |
| Shrews | reading through the bug, doesn't seem like a quick fix | 17:36 |
| JayF | Shrews: I think that was being talked about in here yesterday, the guy who works on our wsme middleware added a patch allows choosing xml/json | 17:36 |
| JayF | but i think the problem was seeing if we could get that in global reqs | 17:36 |
| Shrews | JayF: oh! i missed that | 17:36 |
| JayF | if you log the channel there was a good amount of chatter about it yesterday | 17:36 |
| JayF | and you're better off reading it than getting it through a jay-filter | 17:37 |
| Shrews | /disable filter jay | 17:37 |
| devananda | Shrews: yea. I was just updating that bug | 17:37 |
| devananda | Shrews: the fix in wsme is small, and we could easily patch ironic to remove XML fromthe API ... IF we could require the latest release of WSME | 17:38 |
| devananda | Shrews: which we can't :( | 17:38 |
| devananda | because it's too late in the release cycle | 17:38 |
| devananda | so, we failed to address that in time, and will have to fix it in K | 17:39 |
| Shrews | devananda: ack. thanks for the summary | 17:39 |
| devananda | as a still-incubated project, we can't requset a change to global requirements the day before other priojects cut their RC's | 17:39 |
| devananda | and the fix to WSME hasn't actually landed yet anyway | 17:39 |
| devananda | fwiw: https://review.openstack.org/#/c/123792/ | 17:40 |
| devananda | and https://review.openstack.org/#/c/123520/7 | 17:40 |
| Shrews | oh nice | 17:41 |
| devananda | bug updated | 17:42 |
| devananda | so besides the hash ring bugs, there are only two left | 17:43 |
| devananda | API translation (jroll?) | 17:43 |
| devananda | drivers raising non-ironic exceptions (rloo?) | 17:43 |
| rloo | devananda: in meeting right now. (and i need to listen). in 30 min. | 17:44 |
| devananda | np | 17:46 |
| jroll | hi | 17:48 |
| jroll | devananda: need that done today, I guess? did you decide how large of an impact that is? | 17:49 |
| NobodyCam | hello jroll :> | 17:49 |
| jroll | heya :P | 17:49 |
| *** jcoufal has quit IRC | 17:51 | |
| *** openstackgerrit has quit IRC | 17:51 | |
| *** openstackgerrit has joined #openstack-ironic | 17:53 | |
| devananda | jroll: I agree the impact is relatively minor. we can bump it (and unassign it?) if you're not working on it | 17:57 |
| jroll | devananda: I'm not actively working on it, can be, would prefer not to be right this minute | 17:57 |
| Shrews | devananda: jroll: i can take a stab at that one, but can't promise i can figure it out in time for RC. not sure yet how to fix it. | 17:58 |
| devananda | let's bump it | 17:59 |
| jroll | Shrews: I think the way to go is, enable lazy translation, translate it at the last minute in the error handling middleware | 17:59 |
| devananda | it will be an inconvenience to multi-language teams, but not prevent any functionality | 18:00 |
| NobodyCam | oh jroll did you get the ipa dib element to work? | 18:00 |
| *** todd_dsm has quit IRC | 18:01 | |
| jroll | NobodyCam: no, nothing in the console logs :( | 18:01 |
| NobodyCam | :( | 18:02 |
| *** andreykurilin_ has joined #openstack-ironic | 18:13 | |
| NobodyCam | JayF: question on the "Experimental" tag for drivers idea | 18:16 |
| JayF | NobodyCam: yeah, this is the thing we talked about some at the midcycle | 18:16 |
| JayF | NobodyCam: what's the ? | 18:16 |
| *** jistr has quit IRC | 18:16 | |
| NobodyCam | I really like mordred's idea of positive tags. like "production ready | 18:17 |
| NobodyCam | " | 18:17 |
| NobodyCam | what are your thoughts on switching htat around | 18:17 |
| jroll | ++ | 18:17 |
| jroll | also, we saw how well the experimental tag worked for nova-bm | 18:17 |
| JayF | I think we should be very careful what we name it | 18:18 |
| JayF | but I don't care at all if it's a negative tag or a positive tag | 18:18 |
| JayF | as long as a deployer would have to do an extra step, that's not part of our standard howtos, to use a driver that we generally wouldn't trust | 18:18 |
| jroll | oh, is that the goal? | 18:19 |
| jroll | what's the extra step? | 18:19 |
| NobodyCam | they already have to add the driver to the conf file | 18:19 |
| NobodyCam | to enable it | 18:19 |
| JayF | jroll: use_untested_drivers = true | 18:20 |
| JayF | jroll: or something like that | 18:20 |
| jroll | wat | 18:20 |
| jroll | :/ | 18:20 |
| *** krtaylor has quit IRC | 18:20 | |
| jroll | that's kinda lame | 18:20 |
| JayF | jroll: you're a deployer. Right now how do you know that ipmitool and not impinative is the usual driver people use | 18:20 |
| JayF | fwiw this was talked about at the mid-cycle and everyone seemed to like the idea then :) | 18:21 |
| jroll | well. there's nothing right now that says either of those exist, afaik | 18:21 |
| jroll | other than wiki pages | 18:21 |
| jroll | maaaaaaaybe docs, but I don't think so | 18:21 |
| jroll | so I would try to figure out "what drivers can I use" | 18:21 |
| jroll | google "ironic drivers" | 18:22 |
| jroll | find the wiki page with all drivers, driver status, etc | 18:22 |
| NobodyCam | I think I would rether have non-production ready drivers log a simple "this driver is not tested in production enviroments" | 18:22 |
| jroll | (and testing/CI/deployment info) | 18:22 |
| jroll | NobodyCam: as if anyone reads the conductor startup spam :P | 18:22 |
| NobodyCam | thats another battle | 18:23 |
| NobodyCam | lol | 18:23 |
| JayF | jroll: NobodyCam: I just think about the mysql client option of --safe-updates (it's the default)... what's the opposite option (to disable it)? --i-am-a-dummy | 18:24 |
| jroll | this isn't about "we will slaughter your data" | 18:25 |
| jroll | this is about "this might not work perfect" or "this hasn't been run at scale" | 18:25 |
| *** krtaylor has joined #openstack-ironic | 18:25 | |
| jroll | do we think nova deployers just point at a random driver and decide to use that? | 18:25 |
| jroll | or do they evaluate the list of drivers and try to make an informed decision? | 18:26 |
| NobodyCam | jroll: + the down side if a poor driver is selected by the operator is that they have to change the driver | 18:26 |
| *** zigo has quit IRC | 18:26 | |
| jroll | why should I have to change two config options to use the drac driver? | 18:26 |
| jroll | NobodyCam: right, but that's 10 seconds on the db server | 18:27 |
| JayF | jroll: I think there's a huge difference there; most people will choose their hypervisor before possibly even choosing nova | 18:27 |
| jroll | I'm highly in favor of marking drivers prod ready or not; I don't see the use in making folks take an extra step to use them | 18:27 |
| jroll | JayF: sure | 18:28 |
| JayF | NobodyCam: jroll: No, if a poor drivver is selected by the operator, they'll say "Ironic sucks" and go use something else because we didn't make it clear which ones we know haven't been well tested | 18:28 |
| jroll | there's also the issue of not having much driver overlap | 18:28 |
| JayF | we have tons of driver overlap | 18:28 |
| devananda | JayF: I dont think that's all that different | 18:28 |
| NobodyCam | yep and thats really how I see https://github.com/openstack/ironic/blob/master/etc/ironic/ironic.conf.sample#L220 getting used | 18:28 |
| JayF | especially in power drivers | 18:28 |
| jroll | NobodyCam++ | 18:28 |
| *** zigo has joined #openstack-ironic | 18:28 | |
| jroll | enable prod-ready by default | 18:28 |
| jroll | folks change that as needed | 18:28 |
| devananda | i think operators are likely to choose drivers based on two things | 18:29 |
| devananda | - i already have hardware. does this driver work with it? | 18:29 |
| devananda | - i'm buying new hardware specifically for use with ironic. what does my vendor recommend? | 18:29 |
| devananda | the second case is up to the vendor. hopefully hp, dell, and ibm are giving their customers good advice (and good drivers!) | 18:29 |
| devananda | the first case is where the community is "responsible" | 18:30 |
| devananda | I put that in quotes because we can't dictate what the user/operator will do, but we should give them sane defaults and good suggestions | 18:30 |
| devananda | which I think we do, today | 18:30 |
| devananda | pxe_ipmitool | 18:30 |
| devananda | it's the most tested right now | 18:30 |
| NobodyCam | devananda: ++ yes! | 18:30 |
| JayF | I'm OK with turfing the experimental drivers thing if you guys think that's a better approach | 18:30 |
| devananda | if an operator is going to change the config options, they should do some research into /why/ they're changing them | 18:30 |
| devananda | turfing? | 18:31 |
| NobodyCam | side-lining? | 18:31 |
| devananda | for now, I'd leave them all installed, but not enabled by default | 18:31 |
| JayF | killing, getting rid of, going to the big whiteboard in the sky | 18:31 |
| jroll | JayF: we should still discuss it at the summit | 18:31 |
| devananda | JayF: eh? | 18:31 |
| *** lsmola has quit IRC | 18:31 | |
| jroll | as far as how we want to do it | 18:31 |
| devananda | whcih one(s) are experimental? | 18:31 |
| JayF | turfing == killing the idea :) | 18:31 |
| devananda | *do you see as .. | 18:31 |
| *** lsmola has joined #openstack-ironic | 18:31 | |
| JayF | When we mentioned this at the mid-cycle, we specifically called out the iboot, seamicro, and ipminative drivers | 18:32 |
| JayF | obviously ipminative has CI now so that's a different ball of wax | 18:32 |
| jroll | devananda: I think he's talking about killing that summit session, not the drivers :P | 18:33 |
| devananda | jroll: oh. | 18:33 |
| JayF | Oh yeah that's exactly what I meant | 18:33 |
| JayF | hah | 18:33 |
| devananda | :) | 18:33 |
| * devananda goes back to trying to write PTL email | 18:34 | |
| jroll | lol | 18:34 |
| *** marcoemorais has quit IRC | 18:34 | |
| *** marcoemorais has joined #openstack-ironic | 18:35 | |
| *** marcoemorais has quit IRC | 18:36 | |
| *** marcoemorais has joined #openstack-ironic | 18:37 | |
| NobodyCam | lol... Shrews did you see this: https://github.com/openstack/ironic/blob/master/ironic/db/sqlalchemy/alembic/README#L6-L7 | 18:37 |
| * NobodyCam is now even more confizled by the stamp cmd | 18:37 | |
| *** marcoemorais has quit IRC | 18:37 | |
| *** marcoemorais has joined #openstack-ironic | 18:37 | |
| *** marcoemorais has quit IRC | 18:38 | |
| jroll | seems like stamp is for when you mess up | 18:38 |
| *** marcoemorais has joined #openstack-ironic | 18:38 | |
| rloo | devananda: wrt drivers raising non-ironic exceptions. There are two patches: | 18:40 |
| rloo | devananda: 1. https://review.openstack.org/#/c/122224/. The agent driver. No one has reviewed it recently. | 18:40 |
| rloo | devananda: 2. https://review.openstack.org/#/c/122526/. Changes to conductor. This one needs more discussion. Technically, it isn't part of the bug though. | 18:40 |
| rloo | devananda: so if patch 1 is approved, I would be comfortable saying the bug was addressed. | 18:41 |
| *** pcrews has quit IRC | 18:42 | |
| devananda | rloo: fantastic. thank you for the summary! | 18:46 |
| devananda | jroll: I think 122224 needs your eyes | 18:46 |
| rloo | devananda: yw ;) | 18:46 |
| NobodyCam | devananda: TY for the notes on the hash ring etherpad | 18:47 |
| jroll | rloo, devananda, reviewed, left a -1 | 18:51 |
| jroll | rloo: not sure about my first comment there tbh, it might be valuable | 18:51 |
| * rloo looks | 18:52 | |
| jroll | devananda: folks shouldn't be afraid to review agent code, btw :) | 18:52 |
| *** marcoemorais has quit IRC | 19:00 | |
| *** marcoemorais has joined #openstack-ironic | 19:01 | |
| adam_g | devananda, follow up from meeting: any objection to removing check-tempest-dsvm-ironic-pxe_ssh-nv from stable/icehouse? it hasn't been passing in forever, and is using up precious gate resources | 19:09 |
| devananda | adam_g: i'm sad about that not passing. i thought we had proposed fixes for that? | 19:09 |
| devananda | adam_g: which, btw, I can not approve | 19:09 |
| Shrews | NobodyCam: i did not | 19:10 |
| adam_g | devananda, im not aware of any fixes for it? i dont think tempest folk would be receptive to the kind of special casing we'd need to do to deal with differing APIs between branches | 19:11 |
| *** coolsvap is now known as coolsvap|gone | 19:13 | |
| devananda | adam_g: yea... we would need a feature test for that, and some way to expose it, backported, etc, etc. | 19:14 |
| devananda | adam_g: ok, i'm convinced | 19:14 |
| *** marcoemorais1 has joined #openstack-ironic | 19:14 | |
| adam_g | we could possibly add a periodic job to test the baremetal scenario tests against stable/icehouse, just to have some level of confidence that it still works | 19:14 |
| devananda | adam_g: how are other projects notifying themselves of periodic job results? | 19:15 |
| *** marcoemorais has quit IRC | 19:15 | |
| adam_g | altho, i dont even know if the scenario would pass. IIRC rebuild wasnt fixed until this cycle | 19:15 |
| adam_g | devananda, just email on failures | 19:15 |
| * NobodyCam calls the local dmv and enters the black hole of phone systems | 19:15 | |
| mgagne | Is anyone running Ironic with Nova cells? If yes, at which level are you running Ironic? Api cell or Compute cell? Do you need a dedicated Ironic installation per compute cell? | 19:16 |
| jroll | mgagne: \o | 19:16 |
| jroll | mgagne: here's what we have: | 19:16 |
| mgagne | jroll: (go way!)> \o =) | 19:16 |
| jroll | (as I understand it) | 19:16 |
| mgagne | away* | 19:17 |
| jroll | global cell -> region cell -> ironic cell | 19:17 |
| jroll | it's a nova-api cell | 19:17 |
| mgagne | which one is an api cell ? | 19:17 |
| jroll | and I presume you need an ironic per cell, idk, we're only running it in one cell | 19:17 |
| jroll | the ironic cell | 19:17 |
| jroll | has everything from nova-api to ironic-conductor | 19:17 |
| jroll | (including scheduler, compute, etc) | 19:17 |
| *** ChuckC has quit IRC | 19:18 | |
| *** athomas has quit IRC | 19:19 | |
| mgagne | ok, so what is normally a compute cell is running nova-api and has a dedicated ironic installation. How is nova discovered by Ironic? | 19:19 |
| jroll | not sure if I understand the question, sorry :/ | 19:20 |
| mgagne | and the million dollar question: what about N ironic installations in the same region? | 19:20 |
| jroll | nova is pointed at ironic | 19:20 |
| jroll | we don't have that (yet) but we expect to just spin up identical cells | 19:20 |
| mgagne | jroll: hmm, might be my misunderstanding of Ironic inner working. Nova knows about Ironic but not the other way around right? | 19:21 |
| jroll | correct | 19:21 |
| *** loki184 has quit IRC | 19:21 | |
| mgagne | jroll: how does Nova discovers Ironic? With the catalog or is the API url "hardcoded" in nova.conf? | 19:22 |
| mgagne | jroll: nova.conf: [ironic]/api_endpoint | 19:23 |
| mgagne | jroll: will do my homework first =) | 19:23 |
| jroll | it can be both iirc | 19:23 |
| jroll | s/both/either | 19:23 |
| jroll | actually, it can be either, I fixed a bug with making it work in nova.conf :) | 19:23 |
| jroll | mgagne: ^ | 19:23 |
| devananda | jroll: I don't see any discovery of the endpoint in nova/virt/driver/irnoic/client_wrapper.py | 19:25 |
| jroll | devananda: I think it's in the client | 19:25 |
| * jroll will look in a second | 19:26 | |
| devananda | jroll: the client_wrapper passes ironic_url=CONF.ironic.api_endpoint in every case | 19:26 |
| jroll | oh, does it? huh | 19:26 |
| jroll | ok, thanks for correcting me, I must be thinking of something else | 19:26 |
| devananda | oh - that has no default | 19:26 |
| jroll | maybe ironicclient looks at keystone if it's not provided | 19:27 |
| jroll | or if it's None, I should say | 19:27 |
| devananda | looking in the client ... | 19:27 |
| devananda | 97 endpoint = kwargs.get('ironic_url') or \ | 19:27 |
| devananda | 98 _get_endpoint(_ksclient, **ks_kwargs) | 19:27 |
| jroll | yeah | 19:27 |
| devananda | jroll: so i think you're correct :) | 19:27 |
| mgagne | jroll: so the ironic driver in nova has the ironic api endpoint "hardcoded" (not auto-discovered). http://docs.openstack.org/developer/ironic/deploy/install-guide.html | 19:27 |
| jroll | mgagne: ironicclient will discover it if it is None | 19:28 |
| devananda | mgagne: after looking at the code, i think it will default to autodiscover from keystone (eg, if you dont set the CONF option) | 19:28 |
| jroll | but, I don't trust service catalogs because $reasons | 19:28 |
| devananda | jroll: i want us to default to using keystone for discovering all the endpoints, fwiw | 19:28 |
| devananda | jroll: including glance, swift, etc | 19:28 |
| jroll | devananda: default is fine | 19:28 |
| devananda | jroll: and i fully support you overriding that :) | 19:28 |
| jroll | devananda: I'm sure you've heard about the rackspace service catalog :) | 19:28 |
| jroll | and how great it is | 19:29 |
| mgagne | jroll: where you have to go through hoops to manage to access glance service? =) | 19:29 |
| jroll | mgagne: that's one thing, yeah | 19:30 |
| adam_g | ironic will also use keystoneclient directly to get the ironic API URL for purposes of ramdisk | 19:30 |
| mgagne | jroll: who should I poke to get it fixed? | 19:30 |
| NobodyCam | wow DMV employee was very helpful! /me is shocked | 19:30 |
| jroll | mgagne: to get our service catalog fixed? | 19:30 |
| mgagne | jroll: I suppose | 19:31 |
| jroll | mgagne: probably taylor rhodes at this point | 19:31 |
| jroll | (mostly kidding) | 19:31 |
| jroll | mgagne: we're... working on things | 19:31 |
| jroll | keystone, first | 19:31 |
| mgagne | jroll: is it really called keystone internally? =) | 19:32 |
| russell_h | the thing we run isn't called keystone | 19:32 |
| jroll | mgagne: ha, no, I meant deploying keystone | 19:32 |
| russell_h | but we're deleting that and switching to keystone | 19:32 |
| jroll | rather than what we run | 19:32 |
| mgagne | *wink* *wink* | 19:33 |
| openstackgerrit | Ruby Loo proposed a change to openstack/ironic: Don't reraise Exceptions from agent driver https://review.openstack.org/122224 | 19:37 |
| *** lazy_prince is now known as killer_prince | 19:38 | |
| rloo | jroll: see if I answered your questions/concerns: https://review.openstack.org/122224. | 19:44 |
| *** openstackgerrit has quit IRC | 19:46 | |
| jroll | rloo: +2, thanks | 19:47 |
| rloo | thx jroll! | 19:47 |
| *** openstackgerrit has joined #openstack-ironic | 19:48 | |
| *** pcrews has joined #openstack-ironic | 19:50 | |
| *** coolsvap|gone has quit IRC | 19:51 | |
| *** ChuckC has joined #openstack-ironic | 19:54 | |
| *** marcoemorais1 has quit IRC | 20:04 | |
| *** marcoemorais has joined #openstack-ironic | 20:04 | |
| *** marcoemorais has quit IRC | 20:05 | |
| *** marcoemorais has joined #openstack-ironic | 20:05 | |
| JayF | mgagne: I think our glance api is public? | 20:06 |
| mgagne | jroll: how many cells level do you guys have? I see you mentioning a global cell, would it be like an api cell in IAD? or is the api cell region agnostic? | 20:06 |
| JayF | mgagne: "global" is in a region | 20:06 |
| JayF | mgagne: so we have a "global" nova api in IAD, ORD, DFW, LON, etc | 20:07 |
| JayF | mgagne: OnMetal / Ironic is deployed as one of the cells (alongside normal compute cells) | 20:07 |
| mgagne | JayF: and what jroll referred to as a region cell is in fact a "zone" in the region? | 20:07 |
| jroll | mgagne: oh yeah, listen to JayF, I wasn't braining well when I said that | 20:07 |
| jroll | mgagne: it's two levels, global (region) and onmetal | 20:07 |
| mgagne | JayF: I once managed to authenticate against Glance but now, it's 401 all the time. :-/ | 20:07 |
| JayF | mgagne: if it's publically hittable, it should be somewhat publically usable | 20:08 |
| mgagne | jroll: your brain fart is excused | 20:08 |
| jroll | JayF: our glance doesn't work very well out of the box with standard glance client | 20:08 |
| JayF | ah | 20:08 |
| JayF | I think russell_h might have gotten a working one once though | 20:08 |
| jroll | yeah | 20:08 |
| jroll | so did monty | 20:08 |
| jroll | it took him a whole weekend | 20:08 |
| JayF | We should put that in a devops blog post | 20:09 |
| jroll | (turns out you need to specify the version as v2) | 20:09 |
| JayF | if it's hard to do and we can do it | 20:09 |
| jroll | (and the registry doesn't have the right url) | 20:09 |
| jroll | JayF: maybe we should just make our glance endpoint work like the rest of openstack :/ | 20:09 |
| JayF | jroll: I can't do that; I can write a blog post | 20:09 |
| mgagne | JayF: someone wrote a blog post about it but lost it | 20:10 |
| mgagne | JayF: found it, I used the API key instead of my user password. | 20:15 |
| jroll | yeah, that's another one | 20:16 |
| *** marcoemorais has quit IRC | 20:18 | |
| *** marcoemorais has joined #openstack-ironic | 20:18 | |
| *** marcoemorais has quit IRC | 20:18 | |
| *** marcoemorais has joined #openstack-ironic | 20:18 | |
| * devananda posts nomination to ML | 20:28 | |
| * devananda gets more coffee | 20:28 | |
| NobodyCam | :) | 20:28 |
| jroll | bout time :P | 20:30 |
| devananda | I know :( | 20:35 |
| * rloo wonders what would happen if no one ran for PTL | 20:35 | |
| JayF | last cycle's PTL wins by fiat? | 20:35 |
| jroll | heh | 20:36 |
| jroll | I'm sure someone would swoop in last minute | 20:36 |
| JayF | "You're the PTL again whether you like it or not, mister" | 20:36 |
| jroll | heh | 20:36 |
| devananda | rloo: I don't see any indication of it in http://git.openstack.org/cgit/openstack/governance/tree/reference/charter.rst#n99 | 20:37 |
| devananda | rloo: which probably suggests that it hasn't actually come up before outside of theoretical discussions | 20:37 |
| *** dprince has quit IRC | 20:37 | |
| rloo | ooooo. too bad you submitted something devananda. could have made history... | 20:38 |
| devananda | heh | 20:38 |
| devananda | rloo: i'm OK making other kinds of history :) | 20:38 |
| rloo | :D | 20:38 |
| *** dividehex has quit IRC | 20:44 | |
| NobodyCam | devananda: all kidding aside. that is a very good email ! | 20:45 |
| devananda | ty | 20:45 |
| NobodyCam | no no Thank you :) | 20:45 |
| NobodyCam | brb | 20:45 |
| *** dividehex has joined #openstack-ironic | 20:47 | |
| jroll | +1 | 20:52 |
| * NobodyCam has found a leak in his roof and must go and try to fix... bbiab | 20:57 | |
| *** marcoemorais has quit IRC | 21:06 | |
| *** marcoemorais has joined #openstack-ironic | 21:06 | |
| *** marcoemorais has quit IRC | 21:07 | |
| *** marcoemorais has joined #openstack-ironic | 21:07 | |
| *** marcoemorais has quit IRC | 21:08 | |
| *** marcoemorais has joined #openstack-ironic | 21:10 | |
| *** marcoemorais has quit IRC | 21:10 | |
| *** marcoemorais has joined #openstack-ironic | 21:10 | |
| *** marcoemorais has quit IRC | 21:11 | |
| *** marcoemorais has joined #openstack-ironic | 21:11 | |
| * NobodyCam hopes he has "fixed" his leak | 21:11 | |
| davidlenwell | NobodyCam: is the sky falling ? | 21:14 |
| *** linggao has quit IRC | 21:17 | |
| *** jjohnson2 has quit IRC | 21:25 | |
| NobodyCam | it has been raining a bit here | 21:30 |
| NobodyCam | and hey davidlenwell :) | 21:30 |
| davidlenwell | Hi ! | 21:30 |
| NobodyCam | :) | 21:30 |
| davidlenwell | So we inch closer to having a small cluster at bluebox to donate to the ironic project. | 21:31 |
| jroll | ooo! | 21:31 |
| jroll | what kind of cluster? | 21:31 |
| davidlenwell | something we can use to test and refine the ipmi driver | 21:32 |
| jroll | nice :) | 21:32 |
| davidlenwell | maybe if we do it right use it as an ipmi driver gate test | 21:32 |
| davidlenwell | endless possibilities really | 21:32 |
| jroll | indeed | 21:33 |
| davidlenwell | devananda 's call really | 21:33 |
| jroll | how many nodes in the cluster? | 21:33 |
| * NobodyCam thinks its a cluster of old iphone 5's | 21:33 | |
| davidlenwell | ha ha | 21:33 |
| davidlenwell | lol NobodyCam | 21:33 |
| jroll | lol | 21:33 |
| davidlenwell | we're starting with 5 nodes | 21:33 |
| NobodyCam | :) | 21:33 |
| jroll | only android has ipmi NobodyCam | 21:33 |
| davidlenwell | older hardware | 21:33 |
| * NobodyCam has hte old g1 dev version somewhere | 21:34 | |
| davidlenwell | you're a packrat NobodyCam | 21:34 |
| jroll | davidlenwell: neat, maybe we can test all of the *_ipmitool drivers :) | 21:34 |
| NobodyCam | hehehe /yes | 21:34 |
| davidlenwell | jroll: I think thats the idea | 21:35 |
| jroll | \o/ | 21:35 |
| davidlenwell | from what I am told these 5 nodes are all super micro | 21:35 |
| davidlenwell | but not all the same | 21:35 |
| jroll | should be fine | 21:36 |
| jroll | as long as they work :P | 21:36 |
| davidlenwell | yeah | 21:36 |
| davidlenwell | well idealy they will .. if they don't we'll replace the broken ones .. | 21:36 |
| jroll | idk what bluebox is, but I take it y'all will be deploying ironic eventually? | 21:37 |
| davidlenwell | jroll: thats the idea | 21:37 |
| jroll | woohoo | 21:37 |
| jroll | I like having users | 21:37 |
| davidlenwell | we're working on a proof of concept now | 21:37 |
| davidlenwell | and I've been instructed to start contributing to ironic as well | 21:37 |
| jroll | neat :) | 21:37 |
| NobodyCam | davidlenwell: awesome!!! | 21:37 |
| davidlenwell | So I'll be lurking a bit and trying to find a place I can jump in | 21:38 |
| jroll | cool! | 21:38 |
| NobodyCam | atm DOC's please read our docs as a new user :) | 21:38 |
| davidlenwell | NobodyCam: I can do that | 21:38 |
| jroll | so right now we're about to cut rc1, which means just critical bugs and docs from here to juno final | 21:38 |
| jroll | so yes, read/improve/write docs | 21:38 |
| davidlenwell | do you want me to look at the code docs? | 21:38 |
| rloo | hash rings, hash rings... | 21:39 |
| davidlenwell | that is the docs generated from the doc strings? | 21:39 |
| jroll | davidlenwell: what I would love is for you to work on a test deploy of ironic and tell us if anything is missing from our user/deployer docs | 21:39 |
| jroll | (to start) | 21:39 |
| jroll | and contribute to the missing pieces, of course :P | 21:39 |
| davidlenwell | jroll: we are getting the hardware together for that as well | 21:39 |
| jroll | cool | 21:39 |
| davidlenwell | to do a proof of concept deploy with ironic rather than what we are using now | 21:40 |
| jroll | perfect | 21:40 |
| devananda | davidlenwell: not the inline doc strings. the sphinx docs aimed at operators/deployers | 21:41 |
| davidlenwell | link ? | 21:41 |
| devananda | the things which become this: http://docs.openstack.org/developer/ironic/#admin-guide | 21:41 |
| davidlenwell | okay | 21:41 |
| davidlenwell | are these rst files in the repo? | 21:42 |
| devananda | live here: http://git.openstack.org/cgit/openstack/ironic/tree/doc/source/deploy | 21:42 |
| devananda | yes | 21:42 |
| davidlenwell | ahh .. thanks | 21:42 |
| davidlenwell | thedave lives on! | 21:42 |
| devananda | davidlenwell: I'd say, start by testing out and reviewing the proposed changes | 21:42 |
| devananda | https://review.openstack.org/118614 | 21:42 |
| davidlenwell | devananda: as of right now I don't have real hardware to deploy on .. but I am working on that | 21:43 |
| devananda | Vinay had several other proposed doc changes, i'm not sure why they are abandoned though ... https://review.openstack.org/#/q/owner:%22Vinay+B+S%22,n,z | 21:43 |
| *** marcoemorais has quit IRC | 21:43 | |
| devananda | davidlenwell: ok. well. read away, then test on hardware -- that'd be very helpful | 21:43 |
| davidlenwell | will do | 21:43 |
| JayF | davidlenwell: welcome :) | 21:44 |
| devananda | and, assuming the docs are good, give you an excellent starting point | 21:44 |
| davidlenwell | awesome! | 21:44 |
| JayF | devananda: FWIW; a group of us joined this channel about 9 months ago with no product and a mandate to make one and help upstream... and now some of us are cores and we have a product :) | 21:44 |
| JayF | erm | 21:44 |
| JayF | davidlenwell:^^ | 21:44 |
| JayF | davidlenwell: I wish you the same level of success :) | 21:44 |
| davidlenwell | :) | 21:45 |
| devananda | davidlenwell: where's craig? I don't see him in here yet? | 21:45 |
| davidlenwell | devananda: i nag him | 21:45 |
| ekarlso | JayF: u from hp ? :=) | 21:46 |
| *** harlowja has quit IRC | 21:47 | |
| devananda | ekarlso: nope, he's from rackspace | 21:47 |
| JayF | ekarlso: myself, jroll and a few others in here work on rackspace.com/onmetal | 21:47 |
| *** harlowja has joined #openstack-ironic | 21:47 | |
| devananda | davidlenwell: oh, ignore my second link. it seems Vinay condensed all the other proposals into the virst one I linked, 118614 - so that's the one to read | 21:48 |
| davidlenwell | okay | 21:49 |
| * davidlenwell ignores the second link ;) | 21:49 | |
| *** todd_dsm has joined #openstack-ironic | 21:57 | |
| -openstackstatus- NOTICE: The openstack-infra/config repo will be frozen for project-configuration changes starting at 00:01 UTC. If you have a pending configuration change that has not merged or is not in the queue, please see us in #openstack-infra. | 21:58 | |
| *** harlowja has quit IRC | 22:03 | |
| *** harlowja has joined #openstack-ironic | 22:03 | |
| *** harlowja has quit IRC | 22:09 | |
| *** harlowja_ has joined #openstack-ironic | 22:09 | |
| *** marcoemorais has joined #openstack-ironic | 22:14 | |
| *** todd_dsm has quit IRC | 22:16 | |
| NobodyCam | Shrews: +2'd https://review.openstack.org/#/c/123518 and left a nit | 22:20 |
| *** andreykurilin_ has quit IRC | 22:21 | |
| adam_g | fun race uncovered by the new parallel tempest coverage: https://bugs.launchpad.net/ironic/+bug/1373650 | 22:45 |
| NobodyCam | oh nice... | 22:46 |
| NobodyCam | adam_g: dose it matter for step 3 if that is done in-instance or thru nova reboot | 22:46 |
| NobodyCam | ? | 22:46 |
| *** openstackgerrit has quit IRC | 22:47 | |
| adam_g | NobodyCam, its a test of the nova api | 22:47 |
| adam_g | so, via nova | 22:47 |
| NobodyCam | ack | 22:47 |
| adam_g | NobodyCam, are all drivers power actions processed similarly async? | 22:47 |
| devananda | adam_g: nice! | 22:48 |
| *** openstackgerrit_ has joined #openstack-ironic | 22:48 | |
| devananda | adam_g: so this is similar to another bug that I marked notabug recently, let me find it | 22:48 |
| NobodyCam | yes i believe so | 22:48 |
| *** openstackgerrit_ is now known as openstackgerrit | 22:48 | |
| adam_g | FWIW tempest has config setting 'ready_wait' that can be used to configure an additional sleep after the nova instance has been marked ACTIVE, to actually let the thing boot | 22:49 |
| devananda | adam_g: https://bugs.launchpad.net/nova/+bug/1310135 | 22:52 |
| devananda | one idea i had was setting the power state in Nova right away and _not_ polling ironic for it | 22:53 |
| devananda | that'll be slightly better, but would still race with nova's periodic task | 22:53 |
| devananda | adam_g: the problem with ready_wait here is that a simple time-based wait is prone to still fail when under load, and a condition-based wait may fail (tmieout) if the state changes *before* the condition is checked | 22:55 |
| *** romcheg has quit IRC | 22:58 | |
| devananda | adam_g: if I"m reading nova/compute/manager correctly, the race exists in Nova | 22:58 |
| adam_g | devananda, in the case of the reboot, i think thats essentially what happens. icli.call("node.set_power_state", node.uuid, 'reboot'), ACTIVE | 22:58 |
| devananda | compute manager assumes that a virt driver will return the *new* state immediatey after it requests the change | 22:58 |
| devananda | 2909 self.driver.reboot( | 22:59 |
| devananda | ... | 22:59 |
| devananda | 2935 new_power_state = self._get_power_state(context, instance) | 22:59 |
| devananda | there's no sleep or wait or loopingcall in there | 22:59 |
| devananda | so, because ironic's power state changes are async, it's very likely that Nova will see the instance as ACTIVE at the end of manager.reboot() | 23:00 |
| devananda | the only real solution here is a notification callback from ironic, I think | 23:01 |
| devananda | and having our driver wait for it before returning | 23:01 |
| devananda | at least, unless we look at larger changes in Nova | 23:01 |
| adam_g | yeah, similar to the neutron callbacks | 23:01 |
| devananda | yep | 23:01 |
| adam_g | so a ready_wait will probably avoid the issue in our tempest case, tho would definitely add some minutes to an already long test time | 23:05 |
| * NobodyCam will brb quick walk to store while there is no water falling from the sky | 23:13 | |
| *** igordcard has quit IRC | 23:17 | |
| * adam_g needs to run out as well | 23:22 | |
| *** marcoemorais has quit IRC | 23:27 | |
| *** marcoemorais has joined #openstack-ironic | 23:28 | |
| *** marcoemorais has quit IRC | 23:28 | |
| *** marcoemorais has joined #openstack-ironic | 23:29 | |
| *** marcoemorais has quit IRC | 23:29 | |
| devananda | adam_g: on looking at compute/manager start_instance() and stop_instance() - i'm now of the mind that our nova driver implemented start and stop incorrectly | 23:29 |
| *** marcoemorais has joined #openstack-ironic | 23:29 | |
| devananda | these calls must be blocking within nova.virt.ironic.driver - compute manager assumes that, when the call returns, the operation is complete | 23:30 |
| adam_g | devananda, so ironic driver needs to add loopingcalls between power state changes similar to provision state? | 23:31 |
| devananda | yep | 23:31 |
| devananda | which is easy for on/off | 23:31 |
| devananda | but i'm not sure we can do that for reboot | 23:31 |
| adam_g | use an on / off cycle instead of set_power_state=reboot? :) | 23:32 |
| jroll | devananda: ouch | 23:33 |
| adam_g | need to run out, will recap on backscroll when i return | 23:33 |
| devananda | adam_g: yea, which is going to be slower | 23:33 |
| *** wanyen has quit IRC | 23:34 | |
| * NobodyCam is back | 23:35 | |
| devananda | adam_g: think i have a generic solution. poll for target_power_state != (the one just requested) | 23:37 |
| *** coolsvap has joined #openstack-ironic | 23:39 | |
| *** openstackstatus has quit IRC | 23:55 | |
| *** openstack has joined #openstack-ironic | 23:56 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!