| *** romcheg has joined #openstack-ironic | 00:10 | |
| *** romcheg has left #openstack-ironic | 00:11 | |
| *** radsy has quit IRC | 00:19 | |
| *** matsuhashi has joined #openstack-ironic | 00:26 | |
| *** matsuhashi has quit IRC | 00:33 | |
| *** matsuhashi has joined #openstack-ironic | 00:34 | |
| *** yongli has quit IRC | 00:39 | |
| *** igor_ has joined #openstack-ironic | 00:52 | |
| *** igor_ has quit IRC | 00:57 | |
| *** radsy has joined #openstack-ironic | 01:02 | |
| *** matsuhashi has quit IRC | 01:33 | |
| *** matsuhashi has joined #openstack-ironic | 01:34 | |
| *** nosnos has joined #openstack-ironic | 01:44 | |
| *** igor_ has joined #openstack-ironic | 01:53 | |
| *** igor_ has quit IRC | 01:57 | |
| *** igor_ has joined #openstack-ironic | 02:54 | |
| *** igor_ has quit IRC | 02:58 | |
| *** coolsvap|afk is now known as coolsvap | 03:06 | |
| *** vinbs has joined #openstack-ironic | 03:11 | |
| *** rameshg87 has joined #openstack-ironic | 03:17 | |
| *** nosnos has quit IRC | 03:25 | |
| *** ramineni has joined #openstack-ironic | 03:29 | |
| *** eghobo has joined #openstack-ironic | 03:30 | |
| *** igor_ has joined #openstack-ironic | 03:55 | |
| *** igor_ has quit IRC | 03:59 | |
| *** shausy has joined #openstack-ironic | 04:31 | |
| *** Haomeng has joined #openstack-ironic | 04:35 | |
| *** nosnos has joined #openstack-ironic | 04:46 | |
| *** coolsvap is now known as coolsvap|afk | 04:54 | |
| *** sabah has joined #openstack-ironic | 04:55 | |
| *** igor_ has joined #openstack-ironic | 04:55 | |
| *** igor_ has quit IRC | 05:00 | |
| *** k4n0 has joined #openstack-ironic | 05:03 | |
| *** vkdrao has joined #openstack-ironic | 05:12 | |
| mrda | lucasagomes: we should chat when you're online next re: dell kit and ironic | 05:14 |
|---|---|---|
| *** k4n0 has quit IRC | 05:21 | |
| *** rameshg87 has left #openstack-ironic | 05:24 | |
| *** lazy_prince has joined #openstack-ironic | 05:27 | |
| *** takadayuiko has joined #openstack-ironic | 05:29 | |
| *** k4n0 has joined #openstack-ironic | 05:30 | |
| *** lazy_prince has quit IRC | 05:31 | |
| *** lazy_prince has joined #openstack-ironic | 05:32 | |
| *** eghobo has quit IRC | 05:44 | |
| *** coolsvap|afk is now known as coolsvap | 05:50 | |
| *** rameshg87 has joined #openstack-ironic | 05:50 | |
| *** igor_ has joined #openstack-ironic | 05:56 | |
| *** igor_ has quit IRC | 06:00 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 06:09 | |
| *** sysexit_away has joined #openstack-ironic | 06:10 | |
| *** radsy has quit IRC | 06:13 | |
| *** sysexit_away is now known as sysexit | 06:16 | |
| dtantsur|afk | morning Ironic | 06:30 |
| dtantsur|afk | lifeless, I think we could merge it | 06:31 |
| *** dtantsur|afk is now known as dtantsur | 06:31 | |
| lifeless | dtantsur: you have it marked WIP | 06:35 |
| dtantsur | lifeless, yeah, I'm finishing it right now | 06:35 |
| mrda | hey dtantsur | 06:35 |
| dtantsur | hey mrda | 06:36 |
| mrda | 'grats on the nomination to core! | 06:37 |
| openstackgerrit | lifeless proposed a change to openstack/ironic: Workaround for conflicting volume labels https://review.openstack.org/93133 | 06:37 |
| openstackgerrit | lifeless proposed a change to openstack/ironic: Allow more time for API requests to be completed https://review.openstack.org/93731 | 06:37 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: NodeLocked causing random test failures https://review.openstack.org/95573 | 06:43 |
| dtantsur | mrda, oh, thanks :) | 06:43 |
| dtantsur | lifeless, I've posted what I think could be the final version | 06:44 |
| dtantsur | let us see what Jenkins will say | 06:44 |
| lifeless | cool | 06:46 |
| *** ifarkas has joined #openstack-ironic | 06:50 | |
| *** rameshg87 has left #openstack-ironic | 06:56 | |
| *** igor_ has joined #openstack-ironic | 06:57 | |
| *** igor_ has quit IRC | 07:01 | |
| *** rameshg87 has joined #openstack-ironic | 07:02 | |
| *** rameshg87 has quit IRC | 07:03 | |
| *** boris-42 has quit IRC | 07:03 | |
| *** rameshg87 has joined #openstack-ironic | 07:03 | |
| *** boris-42 has joined #openstack-ironic | 07:11 | |
| Haomeng | dtantsur: 'grats:) | 07:16 |
| *** jcoufal has joined #openstack-ironic | 07:17 | |
| dtantsur | Haomeng, thanks) and good morning | 07:18 |
| *** lazy_prince has quit IRC | 07:19 | |
| Haomeng | dtantsur: :) | 07:19 |
| *** lazy_prince has joined #openstack-ironic | 07:21 | |
| *** pradipta_away is now known as pradipta | 07:28 | |
| *** Haomeng|2 has joined #openstack-ironic | 07:36 | |
| *** Haomeng has quit IRC | 07:37 | |
| openstackgerrit | Yuiko Takada proposed a change to openstack/ironic: Remove "node" parameter from methods handling power state https://review.openstack.org/95670 | 07:51 |
| rameshg87 | hello dtantsur | 08:09 |
| dtantsur | rameshg87, hi | 08:09 |
| *** viktors|afk has quit IRC | 08:09 | |
| rameshg87 | regarding the ilo reviews, we are working on design specs | 08:09 |
| rameshg87 | but should we wait for it before raising code reviews ? | 08:09 |
| rameshg87 | dtantsur: i mean, can we continue the current code reviews ? | 08:10 |
| dtantsur | rameshg87, patch can't be approved until you get your spec approved. Keeping that in mind, we can continue to work on code quality etc | 08:11 |
| rameshg87 | dtantsur - okay, so do you think raising the code review back makes sense ? | 08:11 |
| *** jistr has joined #openstack-ironic | 08:14 | |
| dtantsur | rameshg87, yes, probably. We just won't approve until we have an approved spec (and code clearly corresponds to it). | 08:14 |
| rameshg87 | dtantsur - okay, thanks | 08:14 |
| *** derekh_ has joined #openstack-ironic | 08:17 | |
| *** lucasagomes has joined #openstack-ironic | 08:29 | |
| lucasagomes | morning Ironic :) | 08:33 |
| *** coolsvap is now known as coolsvap|afk | 08:34 | |
| Haomeng|2 | lucasagomes: morning:) | 08:34 |
| mrda | hey lucasagomes | 08:38 |
| mrda | lucasagomes: Are you doing testing on Dell hardware? Can we chat about that briefly? | 08:39 |
| lucasagomes | mrda, hey, I'm not :( I think matt_dubs is | 08:40 |
| takadayuiko | Hi lucasagomes :) | 08:40 |
| mrda | ahh, ok. | 08:40 |
| lucasagomes | mrda, but we can chat anyway | 08:40 |
| lucasagomes | takadayuiko, oh hey! good afternoon there :) | 08:40 |
| mrda | It was mentioned to me that you might be interested in ensuring Ironic works with Dell hardware. | 08:41 |
| mrda | I might need to do similar | 08:41 |
| lucasagomes | mrda, that's correct, RH and Dell has a partnership and we are going to implement a drac driver AFAICT | 08:41 |
| *** igor_ has joined #openstack-ironic | 08:41 | |
| mrda | lucasagomes: do you know what hardware is being targeted? Or the current progress? | 08:42 |
| lucasagomes | mrda, so I think matt_dubs is currently looking at it | 08:42 |
| lucasagomes | mrda, I don't know, but I can ask him today | 08:42 |
| mrda | that's very interesting re: DRAC | 08:42 |
| lucasagomes | mrda, I didn't access the hardware to check it out yet | 08:42 |
| mrda | I'd appreciated that - thank you. I need to go see my family soon at it's been a 12 hour day, and I'll be in trouble soon :) | 08:43 |
| lucasagomes | mrda, yeah, hah I will talk to him today I let you know asap | 08:43 |
| lucasagomes | collaboration is very welcome you guys might want to work together on that | 08:43 |
| mrda | I can probably get some test kit to play with if required. | 08:43 |
| lucasagomes | nice | 08:44 |
| dtantsur | morning lucasagomes! | 08:44 |
| mrda | anyhow, feel free to post here, or PM me, or email michael@the-davies.net once you talk to matty_dubs. | 08:44 |
| mrda | time to go - have a great day ironic! | 08:44 |
| *** mrda is now known as mrda_away | 08:45 | |
| lucasagomes | mrda, ack will do, enjoy ur family! | 08:45 |
| lucasagomes | dtantsur, morning! and congrats :) | 08:45 |
| dtantsur | thanks :) | 08:45 |
| dtantsur | sirushti, hi. Have you seen https://review.openstack.org/#/c/94925/3/specs/juno/better-bootloader-options.rst ? It contains ideas about local boot as well | 08:45 |
| *** coolsvap|afk is now known as coolsvap | 08:45 | |
| takadayuiko | lucasagomes: Yes, it's evening ;) | 08:47 |
| *** nosnos has quit IRC | 08:48 | |
| lucasagomes | takadayuiko, ah, heh good evening then... and how was ur trip back to japan from US? no more jet lag? | 08:48 |
| *** nosnos has joined #openstack-ironic | 08:48 | |
| *** martyntaylor has joined #openstack-ironic | 08:50 | |
| openstackgerrit | Ihar Hrachyshka proposed a change to openstack/ironic: Synced jsonutils from oslo-incubator https://review.openstack.org/95691 | 08:51 |
| takadayuiko | lucasagomes: There is no problem about jet lag, I'm fine! Thank you for concern. But I was in LinuxCon last week, so I couldn't come back to OpenStack :( | 08:52 |
| *** nosnos has quit IRC | 08:53 | |
| lucasagomes | takadayuiko, I see, well that's good as well :D I saw a couple of pictures from the LinuxCon Japan (by greg kroah hartman mostly) last week, really nice! Hope one day I can make it too | 08:53 |
| *** max_lobur has joined #openstack-ironic | 08:55 | |
| *** pradipta is now known as pradipta_away | 08:55 | |
| *** lazy_prince2 has joined #openstack-ironic | 08:59 | |
| *** lazy_prince has quit IRC | 09:00 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic-specs: Add instance_info field to Node model https://review.openstack.org/95701 | 09:02 |
| *** ramineni has left #openstack-ironic | 09:06 | |
| sirushti | dtantsur, Thanks, will do | 09:08 |
| takadayuiko | lucasagomes: yeah, it was my first time for LinuxCon. And it was also my first time seeing "live" Linus! | 09:08 |
| dtantsur | seeing live Linus sounds cool! | 09:08 |
| lucasagomes | takadayuiko, oh haha nice! | 09:08 |
| lucasagomes | I saw him only once as well, it was in 2010 at the first LinuxCon brazil | 09:09 |
| * lucasagomes is getting old :( | 09:09 | |
| sirushti | dtantsur, btw, probably have to factor in EFI boot too. Another related thing to ponder on: having a pxe_config per deploy. I suspect we'd need separate pxe_config templates for different nodes, no? | 09:09 |
| dtantsur | Could you elaborate on this? I'm not quite sure I understand the context | 09:10 |
| *** igordcard has joined #openstack-ironic | 09:10 | |
| *** lazy_prince2 is now known as lazy_prince | 09:11 | |
| lucasagomes | sirushti, for EFI boot we need to create an EFI boot partition and all | 09:11 |
| lucasagomes | I think there's a guy from intel working on that | 09:11 |
| dtantsur | lucasagomes, it may be included into https://review.openstack.org/#/c/94925/3/specs/juno/better-bootloader-options.rst than, what do you think? | 09:12 |
| sirushti | Right, I assumed we'd have separate PXE config files for EFI? | 09:12 |
| lucasagomes | dtantsur, hmm right, well I think it's good to that blueprint make it flexible to extend and add other bootloaders (as needed for UEFI) | 09:13 |
| lucasagomes | but I don't think we should actually do the UEFI work on that blueprint | 09:13 |
| lucasagomes | I would think about it as a plumbing working | 09:13 |
| lucasagomes | sirushti, yes | 09:14 |
| sirushti | lucasagomes, question is, are there plans to have a pxe_config per deploy? | 09:15 |
| sirushti | it'll be a lot more convenient that way right? | 09:16 |
| lucasagomes | sirushti, you mean a pxe config file per machine deployed with the PXE driver? | 09:17 |
| sirushti | lucasagomes, yeah | 09:17 |
| lucasagomes | sirushti, yes, I'm assuming that as well | 09:17 |
| sirushti | dtantsur, lucasagomes thanks for the clarifications :-) | 09:24 |
| lucasagomes | sirushti, np :) | 09:24 |
| *** lokesh184 has joined #openstack-ironic | 09:26 | |
| *** nosnos has joined #openstack-ironic | 09:35 | |
| *** tatyana has joined #openstack-ironic | 09:35 | |
| *** lokesh184 has quit IRC | 09:36 | |
| *** lokesh184 has joined #openstack-ironic | 09:36 | |
| *** lazy_prince2 has joined #openstack-ironic | 09:38 | |
| *** lazy_prince has quit IRC | 09:40 | |
| *** nikunj2512 has joined #openstack-ironic | 09:47 | |
| *** Mikhail_D_ltp has quit IRC | 09:48 | |
| *** matsuhashi has quit IRC | 09:58 | |
| *** matsuhashi has joined #openstack-ironic | 10:00 | |
| *** coolsvap is now known as coolsvap|afk | 10:00 | |
| *** rameshg87 has quit IRC | 10:00 | |
| *** matsuhashi has quit IRC | 10:07 | |
| *** nosnos has quit IRC | 10:07 | |
| *** matsuhashi has joined #openstack-ironic | 10:07 | |
| openstackgerrit | Yuiko Takada proposed a change to openstack/ironic: Remove "node" parameter from methods handling power state in docs https://review.openstack.org/95670 | 10:09 |
| takadayuiko | Tempest patches testing Ironic APIs are: | 10:11 |
| takadayuiko | https://review.openstack.org/#/c/83345/ | 10:11 |
| takadayuiko | https://review.openstack.org/#/c/89472/ | 10:11 |
| takadayuiko | https://review.openstack.org/#/c/86002/ | 10:11 |
| takadayuiko | https://review.openstack.org/#/c/90282/ | 10:11 |
| *** matsuhashi has quit IRC | 10:12 | |
| takadayuiko | We should review these patches as well as possible because we know about Ironic well than Tempest reviewers :) | 10:13 |
| lucasagomes | takadayuiko, :) thank you! I will take a look when I get some free time | 10:13 |
| takadayuiko | lucasagomes: I know you are busy, of course, just FYI. I think that by improving Tempest tests, all of us will be happy. | 10:16 |
| lucasagomes | takadayuiko, +1!! We rely on tempest a lot | 10:18 |
| lucasagomes | everybody should review the ironic related tempest stuff | 10:18 |
| *** rameshg87 has joined #openstack-ironic | 10:20 | |
| *** lazy_prince2 has quit IRC | 10:24 | |
| *** lokesh184 has quit IRC | 10:27 | |
| *** nikunj2512 has quit IRC | 10:27 | |
| *** lazy_prince has joined #openstack-ironic | 10:28 | |
| *** vkdrao has quit IRC | 10:29 | |
| *** nikunj2512 has joined #openstack-ironic | 10:29 | |
| *** lokesh184 has joined #openstack-ironic | 10:43 | |
| dtantsur | vinbs, this bug is supposed to cover your request: https://bugs.launchpad.net/ironic/+bug/1323589 | 10:43 |
| dtantsur | feel free to add comments/ideas/suggestions or just press "Affects me" | 10:44 |
| vinbs | dtantsur, thanks a lot! :) | 10:45 |
| *** rameshg87 has quit IRC | 10:46 | |
| * dtantsur relocates | 10:49 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Nova Ironic Driver to not get deploy r&k from flavor https://review.openstack.org/95551 | 10:50 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Move PXE instance level parameters to instance_info https://review.openstack.org/94855 | 10:50 |
| *** shausy has quit IRC | 11:02 | |
| lucasagomes | dtantsur, thanks for pointing me to tempest, yeah I will have to change it too | 11:02 |
| lucasagomes | :( | 11:02 |
| *** igordcard has quit IRC | 11:02 | |
| *** rameshg87 has joined #openstack-ironic | 11:05 | |
| *** rameshg87 has left #openstack-ironic | 11:17 | |
| *** lokesh184 has quit IRC | 11:19 | |
| *** nikunj2512 has quit IRC | 11:22 | |
| *** romcheg has joined #openstack-ironic | 11:24 | |
| *** sabah has quit IRC | 11:29 | |
| openstackgerrit | Yuiko Takada proposed a change to openstack/ironic: Add util method generating random uuid https://review.openstack.org/95733 | 11:42 |
| *** Alexei_9871 has joined #openstack-ironic | 11:44 | |
| *** lucasagomes is now known as lucas-hungry | 11:54 | |
| * lucas-hungry lunch time! | 11:54 | |
| romcheg | Morning guys! | 11:56 |
| dtantsur | morning romcheg! | 12:00 |
| takadayuiko | Hi, romcheg | 12:01 |
| takadayuiko | Good bye everyone, I'll go home! | 12:02 |
| romcheg | How're you guys doing? | 12:02 |
| *** takadayuiko has quit IRC | 12:02 | |
| romcheg | Bye takadayuiko! | 12:02 |
| dtantsur | Folks, do we have a way of marking a bug as `probably requiring a spec`? | 12:02 |
| dtantsur | First of all, I don't think they should be "Triaged", as w/o a spec it's not obvious how to fix them | 12:03 |
| NobodyCam | good morning Ironic... | 12:16 |
| dtantsur | morning NobodyCam! | 12:17 |
| NobodyCam | dtantsur: we could add a commment saying I think this change requiers a spec | 12:17 |
| NobodyCam | and morning dtantsur | 12:18 |
| NobodyCam | oh internets are slow this morning | 12:18 |
| dtantsur | Also re https://review.openstack.org/#/c/91585/ : how many pages are we actually expecting when paginating? | 12:20 |
| dtantsur | (that's mostly for lucas-hungry, but others can think as well) | 12:20 |
| *** linggao has joined #openstack-ironic | 12:20 | |
| NobodyCam | ahh :) i need to get with lucas-hungry this morning | 12:20 |
| NobodyCam | but first I need to wake up :-p | 12:22 |
| NobodyCam | brb morning walkies time | 12:22 |
| *** jdob has joined #openstack-ironic | 12:33 | |
| *** rloo has joined #openstack-ironic | 12:37 | |
| dtantsur | NobodyCam, or someone: could you set status of this https://bugs.launchpad.net/python-ironicclient/+bug/1229324 to "Fix Release". I'm getting "Forbidden" on it | 12:40 |
| *** k4n0 has quit IRC | 12:40 | |
| *** rloo has quit IRC | 12:42 | |
| *** rloo has joined #openstack-ironic | 12:42 | |
| NobodyCam | dtantsur: was the patch tagged with that bug? | 12:43 |
| dtantsur | NobodyCam, seems like yes | 12:43 |
| dtantsur | sometimes it does no work. Also it does not work for Partial-Bug | 12:43 |
| NobodyCam | ya | 12:44 |
| openstackgerrit | Dan Prince proposed a change to openstack/ironic: Add parallel_image_downloads option https://review.openstack.org/95213 | 12:49 |
| *** vinbs_ has joined #openstack-ironic | 12:50 | |
| *** rloo has quit IRC | 12:51 | |
| *** romcheg has quit IRC | 12:52 | |
| *** rloo has joined #openstack-ironic | 12:52 | |
| *** romcheg has joined #openstack-ironic | 12:52 | |
| dtantsur | devananda, when you get here: are there any reasons now to keep https://review.openstack.org/#/c/73956/ blocked? | 12:52 |
| *** vinbs has quit IRC | 12:52 | |
| Shrews | morning all | 12:52 |
| NobodyCam | good morning Shrews | 12:52 |
| *** jgrimm has quit IRC | 12:53 | |
| *** vinbs_ has quit IRC | 12:56 | |
| *** jistr has quit IRC | 12:57 | |
| *** vinbs has joined #openstack-ironic | 12:59 | |
| *** sseago__ has joined #openstack-ironic | 13:01 | |
| *** sseago_ has quit IRC | 13:01 | |
| dtantsur | Shrews, morning | 13:02 |
| *** jistr has joined #openstack-ironic | 13:03 | |
| *** sysexit has quit IRC | 13:03 | |
| rloo | NobodyCam: morning! (Shrews, dtantsur, and all ironic'ers too) | 13:03 |
| dtantsur | morning rloo :) | 13:04 |
| rloo | NobodyCam: wrt 93133... guess you're going to rework it? | 13:04 |
| NobodyCam | rloo yep... working it now.. | 13:04 |
| NobodyCam | and good morning | 13:05 |
| * dtantsur is pissed off by Launchpad :-[ | 13:06 | |
| *** lucas-hungry is now known as lucasagomes | 13:07 | |
| lucasagomes | morning NobodyCam romcheg rloo | 13:07 |
| NobodyCam | morning lucasagomes | 13:07 |
| lucasagomes | dtantsur, by default I think we have 1000 items per page | 13:07 |
| NobodyCam | hey hey lucasagomes gots a second | 13:07 |
| *** jcoufal has quit IRC | 13:07 | |
| *** rloo has quit IRC | 13:07 | |
| lucasagomes | so idk if we are going to expect 1000 pages | 13:07 |
| *** jcoufal_ has joined #openstack-ironic | 13:07 | |
| lucasagomes | and 1000000 items registered | 13:07 |
| *** rloo has joined #openstack-ironic | 13:08 | |
| dtantsur | lucasagomes, if we never expect >500 pages, I think I'm ok with that patch :) | 13:08 |
| dtantsur | NobodyCam, could you close this https://bugs.launchpad.net/python-ironicclient/+bug/1285530 as invalid? Appropriate patch got -2 | 13:08 |
| lucasagomes | dtantsur, yeah, well I don;t want to set a limit | 13:08 |
| lucasagomes | lemme look at the code see the flaws | 13:08 |
| dtantsur | lucasagomes, recursion. Python limit is 1000 | 13:08 |
| lucasagomes | right | 13:09 |
| * dtantsur is not quite sure that generators do involve recursion here | 13:09 | |
| dtantsur | but I'd prefer to be on a safe side | 13:09 |
| lucasagomes | yeah that's what I'm thinking because I'm using yield there | 13:10 |
| lucasagomes | idk if we are going to hit that problem | 13:10 |
| lucasagomes | I will make a test | 13:10 |
| *** jbjohnso has joined #openstack-ironic | 13:11 | |
| dtantsur | lucasagomes, yeah, test is the best way to decide | 13:11 |
| NobodyCam | lucasagomes: have you seen lifeless comments on 93133, just wanted you to see his comments | 13:13 |
| lucasagomes | NobodyCam, will take a look | 13:13 |
| NobodyCam | I about to rework deploy_utils to not repartition on rebuilds | 13:14 |
| NobodyCam | lucasagomes: his comments are on rev 35 | 13:15 |
| lucasagomes | NobodyCam, to not repartiton!? | 13:15 |
| lucasagomes | sometimes you need it | 13:15 |
| NobodyCam | thats why I wanted to chat about it | 13:16 |
| lucasagomes | if the new image is bigger than the previous one (and no grow rootfs utility were used) | 13:16 |
| lucasagomes | right | 13:16 |
| lucasagomes | I will check his comments, but I don't think it's a safe assumption that we don't need to repartition | 13:16 |
| *** praefect has joined #openstack-ironic | 13:17 | |
| NobodyCam | if new image is larger and preserve_ephemeral is set. we could end up trashing the ephemeral | 13:17 |
| *** sysexit has joined #openstack-ironic | 13:19 | |
| lucasagomes | NobodyCam, no, because the root is the last partition | 13:19 |
| lucasagomes | the ephemeral is the first | 13:19 |
| lucasagomes | so it won't affect it | 13:19 |
| NobodyCam | oh thats right | 13:19 |
| *** dtantsur has quit IRC | 13:32 | |
| *** rakesh_hs has joined #openstack-ironic | 13:33 | |
| *** matty_dubs|gone is now known as matty_dubs | 13:39 | |
| *** dtantsur has joined #openstack-ironic | 13:41 | |
| NobodyCam | brb | 13:42 |
| lucasagomes | NobodyCam, commented on patch #35 | 13:44 |
| lucasagomes | NobodyCam, I don't I feel that we are over complicating it | 13:44 |
| lucasagomes | NobodyCam, we shouldn't be wipping data from the disk like that or caring whether it's MBR GPT etc... | 13:45 |
| lucasagomes | NobodyCam, IMO if it's a problem with parted for not being less error-prone, maybe we should just use sfdisk/sgdisk in the background to do the partitioning | 13:46 |
| lucasagomes | and conserve the same interface for the DiskPartitioner (add_partition, get_partitions, commit) | 13:47 |
| *** jgrimm has joined #openstack-ironic | 13:47 | |
| lucasagomes | which seems to be generic enough | 13:47 |
| dtantsur | may make sense | 13:48 |
| dtantsur | also it can change situation around last 1MiB | 13:48 |
| lucasagomes | (and sfdisk friendly, since parted use start/end to create the partitions, and sfdisk uses size, and the DiskPartitioner interface expect size as a parameter to add a partition) | 13:48 |
| lucasagomes | yeah | 13:48 |
| lucasagomes | s/I don't feel/I _do_ feel/g | 13:49 |
| *** jcoufal has joined #openstack-ironic | 13:49 | |
| NobodyCam | reading your comments now | 13:51 |
| *** jcoufal_ has quit IRC | 13:52 | |
| *** vinbs has quit IRC | 13:52 | |
| *** jistr has quit IRC | 13:53 | |
| openstackgerrit | Ruby Loo proposed a change to openstack/ironic: Implement API to get driver properties https://review.openstack.org/73005 | 13:53 |
| *** igor_ has quit IRC | 13:54 | |
| NobodyCam | lucasagomes: if I read your comments correctly I need to remove parted and insert sfdisk? | 13:56 |
| lucasagomes | NobodyCam, well, I would say maybe not we can just easily workaround this problem by removing the labels and then adding it back in the past | 13:58 |
| lucasagomes | but idk I think that the time we spent talking about this workaround | 13:58 |
| lucasagomes | is not worth | 13:58 |
| lucasagomes | if we are going to spent time thinking about a workaround it would be better to just do the right thing | 13:58 |
| lucasagomes | and not having a workaround at all | 13:58 |
| lucasagomes | if sfdisk and sgdisk is more resilient to such errors with the labels, grand, let's just it | 13:59 |
| *** annegentle has joined #openstack-ironic | 13:59 | |
| NobodyCam | i can agree I never saw this growing to 35 + revs, | 14:00 |
| lucasagomes | NobodyCam, looking at the DiskPatitioner interface, I don't think that we are going to need many (if any) changes at it | 14:00 |
| lucasagomes | NobodyCam, yeah | 14:00 |
| lucasagomes | I was ok in just removing the labels with tune2fs | 14:00 |
| lucasagomes | but now it's total new level of complexity we are introducing there that might not worth to spent time in that workaround | 14:01 |
| NobodyCam | the removing labels was what it started as, and that got shot down | 14:01 |
| lucasagomes | yeah | 14:01 |
| rloo | lucasagomes, NobodyCam: me confused. How did it go from dealing with the labels, to wiping the start/end of the disk? | 14:01 |
| lucasagomes | I know | 14:01 |
| lucasagomes | rloo, see | 14:01 |
| lucasagomes | that's the thing | 14:01 |
| lucasagomes | we are not even thinking about the initial problem anymore | 14:01 |
| lucasagomes | the initial problem is the labels | 14:01 |
| rloo | lucasagomes, NobodyCam: I got the feeling that the wiping was needed to address a different issue. | 14:01 |
| lucasagomes | we could simply iteract over the partitions and issue a tune2fs to remove the label | 14:01 |
| dtantsur | I also have no idea, why tune2fs won't work | 14:02 |
| lucasagomes | rloo, we are not :( | 14:02 |
| lucasagomes | dtantsur, I think it does | 14:02 |
| lucasagomes | NobodyCam, it does right? | 14:02 |
| lucasagomes | rloo, but wipping disks is a general problem for ironic | 14:02 |
| lucasagomes | rloo, we should wipe the whole data of the disk when a new tenant will use the machine... but anyway it's a diff problem and can't be done via iscsi | 14:03 |
| NobodyCam | this was rev 2. which passed my local rebuild tests: https://review.openstack.org/#/c/93133/2/ironic/drivers/modules/deploy_utils.py | 14:03 |
| NobodyCam | i just want to work on some thing that has a chance of landing | 14:04 |
| *** jistr has joined #openstack-ironic | 14:04 | |
| openstackgerrit | Rakesh H S proposed a change to openstack/ironic: ipmi double bridging functionality https://review.openstack.org/95775 | 14:04 |
| dtantsur | NobodyCam, I like rev 2 except for absent logging | 14:04 |
| lucasagomes | lifeless, ping re parted problem with labels | 14:04 |
| NobodyCam | :) | 14:04 |
| lucasagomes | NobodyCam, I feel sorry cause u've been very active on that problem solving it diff ways, and now I'm here saying to use sfdisk | 14:05 |
| lucasagomes | but IMO, right now I think that's the best way, simply because that won't require a workaround | 14:05 |
| rloo | lucasagomes, NobodyCam: yeah, best to all agree. Sorry, I guess I should have asked initially why you were wiping out the disk etc but I assumed it was needed. | 14:05 |
| NobodyCam | rloo: wipping may still be required for the reasions lifeless pointed out. removing raid / lvm data | 14:06 |
| rloo | NobodyCam: yes, that would be a different bug though. | 14:07 |
| dtantsur | Why strip lvm data btw? | 14:09 |
| NobodyCam | dtantsur: when we initially partition a disk we need to ensure it is like a new disk. we can not antisapate the state we get disks in | 14:11 |
| NobodyCam | * anticipate | 14:12 |
| dtantsur | NobodyCam, I thought making a new partition will cancel any previous state (except for labels bug) | 14:12 |
| dtantsur | if we can simulate things on file, we can write a test and see | 14:12 |
| NobodyCam | dtantsur: some raids write data to the disk | 14:12 |
| dtantsur | NobodyCam, raids are even harder, I guess. Do we want to wipe their data or just the opposite? | 14:13 |
| NobodyCam | dtantsur: on a new deploy yes we do want to wipe. | 14:14 |
| NobodyCam | if its a rebuild maybe not so much | 14:14 |
| dtantsur | NobodyCam, what's gonna happen to raid than? | 14:14 |
| rloo | dtantsur: wrt bug 1323589, lack of docn, and the etherpad: https://etherpad.openstack.org/p/Using_Ironic_to_launch_instances_on_real_hardware | 14:14 |
| rloo | dtantsur: the link for 2) in the etherpad is old. The page got moved to a diff url, but I guess we still keep the old page. | 14:15 |
| NobodyCam | dtantsur: if its a new deploy I would not want left over raid config from last tenant on my disk | 14:16 |
| dtantsur | rloo, it's not mine etherpad actually, but will fix, thank you | 14:16 |
| rloo | dtantsur: I just updated it ;) | 14:16 |
| rloo | dtantsur: didn't know it wasn't yours. | 14:17 |
| dtantsur | rloo, good! I had a strange feeling right now, trying to understand what is wrong :D | 14:17 |
| rloo | dtantsur: ha ha. Wonder how we can get the old page deleted. | 14:17 |
| NobodyCam | bbt ... brb | 14:19 |
| dtantsur | rloo, so, page http://docs.openstack.org/developer/ironic/deploy/install-guide.html is mostly what we actually were looking for in https://bugs.launchpad.net/ironic/+bug/1323589 | 14:19 |
| dtantsur | maybe need some updates, right? | 14:19 |
| rloo | dtantsur: ah, I was wondering, which is why I looked. but then I saw the etherpad, and I wondered. | 14:20 |
| rloo | dtantsur: guess I do a lot of 'wonder'ing. | 14:20 |
| dtantsur | rloo :) thanks for pointing, I'll check, what can be fixed for that bug | 14:20 |
| rloo | dtantsur: yeah, if that page works for you, we could enhance it. The idea is that whenever Ironic graduates, it'll have 'official' install docs. | 14:20 |
| rloo | dtantsur: I even think it might have been you that mentioned adding tftp stuff in that doc a few months ago. I think I ran out of steam and wasn't sure how much detail to add to it. | 14:21 |
| rloo | dtantsur: sorry, not you. It was ALeksandr Gordeev. I have some notes about docn, in an etherpad: https://etherpad.openstack.org/p/IronicDocumentationTasks | 14:22 |
| matty_dubs | Hmm, this is reminding me that I've been failing to do much with docs... | 14:25 |
| matty_dubs | :'( | 14:25 |
| NobodyCam | morning matty_dubs | 14:25 |
| matty_dubs | Morning NobodyCam | 14:25 |
| dtantsur | rloo, how do you like it now: https://bugs.launchpad.net/ironic/+bug/1323589 ? | 14:26 |
| dtantsur | morning, matty_dubs | 14:26 |
| rloo | morning matty_dubs! | 14:26 |
| rloo | dtantsur: I wonder if there is too much in this one bug. eg, is fedora even ready? i haven't looked recently to see if ironic is packaged for fedora. | 14:27 |
| lucasagomes | matty_dubs, morning | 14:28 |
| rloo | dtantsur: maybe use the etherpad to track. or a blueprint. and separate bugs for each part or something? | 14:28 |
| dtantsur | rloo, it is, at least in testing repo: http://repos.fedorapeople.org/repos/openstack-m/openstack-m/fedora-20/noarch/ | 14:28 |
| matty_dubs | Howdy rloo, lucasagomes ! | 14:28 |
| dtantsur | rloo, anyway, it's an optional part of the bug | 14:28 |
| dtantsur | I'm running Ironic and DevStack on Fedora as well | 14:29 |
| rloo | dtantsur: so it is fine. as long as I'm not updating the docn :-) | 14:29 |
| dtantsur | :) | 14:29 |
| rloo | dtantsur: sorry, the migration from nova baremetal is a totally different beast. i'd remove that! | 14:30 |
| dtantsur | sure, as you wish | 14:31 |
| dtantsur | rloo, as author of https://bugs.launchpad.net/python-ironicclient/+bug/1304488 could you give your opinion on the discussion here: https://review.openstack.org/#/c/65865/ ? | 14:31 |
| rloo | dtantsur: sure... | 14:32 |
| rloo | dtantsur: do you know why we need/want argcomplete? or do i have to wade through it all to find out? | 14:33 |
| dtantsur | rloo, patch author point is that it provides correct autocompletion for subcommands | 14:33 |
| dtantsur | rloo, you can see what he writes in comments - I didn't check myself | 14:34 |
| rloo | I only added bug 1304488 cuz the doc folks have a script that uses that to generate docn. | 14:34 |
| rloo | if the doc folks aren't going to update their script to use argcomplete, i don't see any reason to have it. wish there was more context to these 'dropped' patches. | 14:35 |
| dtantsur | rloo, honestly, I would prefer to fix existing solution, rather than introducing yet another | 14:39 |
| rloo | dtantsur: sigh. trying to think how to word this politely. | 14:39 |
| rloo | dtantsur: do you think it is acceptable for people to review something, w/o knowing why we want that thing in the first place? | 14:40 |
| rloo | dtantsur: you already asked about a bug, and the relationship to bash-completion. I am going to -1 it cuz there's no bug and no rationale for why I'd want this. | 14:41 |
| dtantsur | rloo, that makes sense. While I understand and appreciate good intention of the author, we need some order in what, when and how we add to Ironic | 14:42 |
| dtantsur | For fixing completion (if it's really broken, that is), I would start with posting to ML and talking to all commands | 14:43 |
| rloo | dtantsur: yup. | 14:43 |
| *** rakesh_hs has quit IRC | 14:51 | |
| rloo | matty_dubs: hey, wrt https://review.openstack.org/#/c/83577/. Are you OK if I -2 it? Or is 'abandoned' sufficient? | 14:54 |
| matty_dubs | -2 is fine by me. I'm not sure how to withdraw a patch | 14:54 |
| rloo | matty_dubs: ok. (I don't know either.) | 14:55 |
| *** romcheg has left #openstack-ironic | 14:58 | |
| matty_dubs | rloo: Some of us were joking in Atlanta that in addition to the '+2' dinner for cores, we should have a separate party for anyone who's received a -2. I'm in! ;) | 15:00 |
| rloo | matty_dubs: I think that if you haven't received a -2, you haven't lived yet ;) | 15:01 |
| matty_dubs | Haha, that too | 15:01 |
| lucasagomes | hah | 15:04 |
| lucasagomes | that would be a great party | 15:04 |
| lucasagomes | I could as well :) | 15:04 |
| lucasagomes | could go* | 15:04 |
| *** blamar has quit IRC | 15:06 | |
| linggao | ping devananda, | 15:10 |
| *** blamar has joined #openstack-ironic | 15:12 | |
| *** BadCub has joined #openstack-ironic | 15:13 | |
| NobodyCam | brb... maken coffee.. quick walkies | 15:17 |
| *** romcheg has joined #openstack-ironic | 15:18 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Add some real-world testing on DiskPartitioner https://review.openstack.org/94620 | 15:21 |
| dtantsur | NobodyCam, rloo may one of you finish reviewing https://review.openstack.org/#/c/92625/ please? It's still about caching in PXE | 15:22 |
| rloo | dtantsur: are you reading my mind? I just started looking at that ;) | 15:23 |
| dtantsur | lol | 15:23 |
| * dtantsur has some errands, be back in ~2 hours | 15:25 | |
| *** dtantsur is now known as dtantsur|afk | 15:25 | |
| *** mkerrin1 has quit IRC | 15:26 | |
| *** lazy_prince has quit IRC | 15:26 | |
| *** mkerrin has joined #openstack-ironic | 15:30 | |
| Shrews | dtantsur|afk: rloo: there seems to be an 'else' associated with a 'for' in that review. indentation fail, me thinks :) | 15:31 |
| rloo | Shrews: you are faster than me. I haven't gotten (or I missed) that. | 15:32 |
| Shrews | modules/pxe.py | 15:32 |
| NobodyCam | lucasagomes: looking at nova baremetal we had this: https://github.com/openstack/nova/blob/master/nova/cmd/baremetal_deploy_helper.py#L105 | 15:33 |
| rloo | Shrews: I see it. | 15:34 |
| NobodyCam | think thats still needed? | 15:34 |
| lucasagomes | NobodyCam, not that I know | 15:34 |
| lucasagomes | I mean, idk why it was added in the first place | 15:34 |
| rloo | Shrews: yeah, needs to be out-dented or whatever it is called. | 15:34 |
| Shrews | rloo: yep | 15:35 |
| dtantsur|afk | Shrews, rloo folks, for..else.. is a valid construction | 15:36 |
| dtantsur|afk | Shrews, rloo else block is executed, if for loop was not intervented using `break` | 15:36 |
| dtantsur|afk | Shrews, rloo that's exactly what I meant: if `break` was never hit, raise an exception | 15:37 |
| rloo | dtantsur|afk: oh, interesting. | 15:37 |
| dtantsur|afk | now I'm really afk, thank you for reviewing :) | 15:38 |
| Shrews | well that's an odd construct | 15:38 |
| Shrews | will re-review with that in mind | 15:39 |
| *** jgrimm has quit IRC | 15:39 | |
| *** graf_Pogonialo has joined #openstack-ironic | 15:42 | |
| *** graf_Pogonialo has left #openstack-ironic | 15:42 | |
| *** eghobo has joined #openstack-ironic | 15:43 | |
| *** jgrimm has joined #openstack-ironic | 15:52 | |
| *** shakamunyi has joined #openstack-ironic | 15:56 | |
| NobodyCam | doh...No valid host was found. Reason: No conductor service registered which supports driver pxe_ssh. (HTTP 400) | 15:58 |
| NobodyCam | time to rebuild | 15:58 |
| devananda | morning, all | 15:58 |
| NobodyCam | good morning devananda :) | 15:59 |
| lucasagomes | morning devananda | 15:59 |
| * devananda catches up on scrollback | 15:59 | |
| lucasagomes | devananda, quick q... re instance_info spec... you mentioned to create a migration script | 15:59 |
| lucasagomes | is it like a bash script? something like pass the flavor id and the node id, so it extracts the deploy k&r parameters from the flavor and set it to the node's instance_info field | 16:00 |
| lucasagomes | ? | 16:00 |
| devananda | lucasagomes: yea, something for an existing deployment (eg, of icehouse) | 16:01 |
| devananda | lucasagomes: we need to support existing users' upgrade path | 16:01 |
| lucasagomes | devananda, gotcha, cool will do it | 16:01 |
| lucasagomes | yeah I was just a bit confused whether it would be a bash script or something else | 16:01 |
| lucasagomes | like a step to step instruction etc | 16:01 |
| *** matty_dubs is now known as matty_dubs|lunch | 16:03 | |
| devananda | reviewing 95573 now so we can get this fixed for CI | 16:04 |
| lucasagomes | cool, yeah i added u to the review list cause ur input is important there | 16:05 |
| *** vinbs has joined #openstack-ironic | 16:06 | |
| devananda | sirushti, lucasagomes: I don't see a reason for per-node pxe configs yet. OTOH, we do need more than one pxe config per conductor | 16:07 |
| devananda | sirushti, lucasagomes: so AIUI, we should provide one pxe config per /type/ of deploy, and link nodes as needed to that config | 16:08 |
| devananda | ooh -- nice tepest stuff! https://review.openstack.org/#/q/status:open+project:openstack/tempest+branch:master+topic:bp/missing-baremetal-api-test,n,z <-- adam_g, you may be interested | 16:09 |
| lucasagomes | devananda, I might be confused here... but we create one pxe config per node, using the mac address of that node | 16:10 |
| lucasagomes | or am I mixing stuff? | 16:10 |
| devananda | lucasagomes: but using a single template that only supports PXE booting a linux kernel | 16:11 |
| devananda | lucasagomes: AIUI, we need to support different config templates for windows, iPXE, and EFI | 16:11 |
| lucasagomes | right, sure yeah | 16:11 |
| lucasagomes | this could be an parameter to the node, and if not present we fallback to the default config | 16:12 |
| devananda | dtantsur|afk: re: editing status of bug 1229324 -- LP is broken on bugs which were filed like that, and it's not possible for anyone to update them via the UI | 16:15 |
| lucasagomes | re tempest, here's another one: https://review.openstack.org/#/c/95789/ , needed by the instance_info stuff | 16:15 |
| devananda | dtantsur|afk: there is apparently an email interface for LP which works, some times, for those bugs.... but i haven't made the time to figure it out yet | 16:15 |
| devananda | dtantsur|afk: also I unblocked 73956. thanks for the ping! | 16:16 |
| *** tatyana has quit IRC | 16:17 | |
| Shrews | lucasagomes: fyi, that tempest change may conflict with mine: https://review.openstack.org/94436 | 16:18 |
| Shrews | lucasagomes: mainly b/c i fixed the class name | 16:19 |
| devananda | NobodyCam, lucasagomes, rloo: on the disk partitioner patch set -- why are we talking about wiping lvm data when we don't support creating it? | 16:20 |
| NobodyCam | devananda: it was lifeless concern | 16:21 |
| devananda | i've been pretty clear in my conversations with the mirantis Fuel team that, at least for this cycle, I don't want Ironic to support LVM or software RAID | 16:21 |
| NobodyCam | we do not know what the disk had before our first oartition | 16:21 |
| NobodyCam | partition even | 16:21 |
| NobodyCam | not but if the disk has lvm or raid on it we should wipe it no? | 16:22 |
| lucasagomes | Shrews, hmm right... I can add a dependency on urs if needed | 16:22 |
| lucasagomes | but maybe git will be smart enough because the function I changed you didn't touch in ur patch, which is validate_driver_info() | 16:22 |
| Shrews | lucasagomes: it might merge ok. just leave it for now | 16:22 |
| devananda | rloo, dtantsur|afk: regarding official install docs -- that page (developer/ironic/deploy/install-guide) is gonna be it for a while, I think. | 16:23 |
| lucasagomes | devananda, +1, we probably don't need to wipe _any_ data | 16:23 |
| Shrews | i was just giving a heads-up | 16:23 |
| lucasagomes | Shrews, I see, thanks for that | 16:23 |
| rloo | devananda: wrt docn. Yeah, until end of Juno at the earliest :-( | 16:24 |
| devananda | rloo: I wouldn't expect it until 6 months after graduation TBH | 16:24 |
| rloo | devananda: really? I thought it was a requirement *for* graduation. Wow. | 16:24 |
| devananda | rloo: I found out that what we have is what's required for graduation -- not inclusion in the openstack-docs repo | 16:25 |
| rloo | devananda: ahh. ok. We'll just try to keep our version up-to-date then... | 16:25 |
| *** derekh_ has quit IRC | 16:26 | |
| devananda | linggao: pong | 16:27 |
| rloo | devananda: btw, wrt docn, if there is no official docn til later, do we need to use the DocImpact flag? | 16:28 |
| linggao | devananda, first we are going to setup a third party CI for the ipminative deriver in a few weeks. We are preparing it now. | 16:29 |
| rloo | devananda: https://wiki.openstack.org/wiki/Documentation/DocImpact. Anne was surprised that I didn't know about it ;) | 16:29 |
| *** jistr has quit IRC | 16:30 | |
| linggao | devananda, I am looking at what features I can contribue for Juno relase. There are a few things come to my mind. Just want to run through you before I dig into details for the blueprints. | 16:30 |
| linggao | 1. intergrate console with horizon | 16:31 |
| linggao | 2. intergrate console with nova command line. | 16:31 |
| linggao | 3. add ipminative console support. | 16:31 |
| devananda | rloo: true - we don't need to use it, but it's helpful for us as well, isn't it? | 16:31 |
| *** jcoufal has quit IRC | 16:32 | |
| devananda | rloo: perhaps we should set up a gerrit trigger that files a bug against ironic so we know to come back and udpate the docs ;) | 16:32 |
| linggao | 4. reset service processor (BMC or bldes's on-board service processor) | 16:32 |
| linggao | devananda, what do you think? | 16:32 |
| rloo | devananda: it supposedly sends email to openstack-docs@. If they're OK with that, then yes, we ought to 'get into the habit'. Presumably, the blueprint specs will indicate it too. | 16:33 |
| devananda | linggao: re: CI for ipminative -- that's fantastic! there is a new weekly meeting that anteaya started, specifically for third-party-CI. you may be interested in following that | 16:33 |
| linggao | devananda, that's great. I'll contact anteaya. | 16:34 |
| devananda | linggao: getting console log (rather than interactive console) would be very helpful. this should be exposed also via Nova CLI and via Horizon (it should plug in fairly easily) | 16:34 |
| anteaya | o/ | 16:34 |
| NobodyCam | morning anteaya :) | 16:34 |
| anteaya | https://wiki.openstack.org/wiki/Meetings/ThirdParty | 16:34 |
| devananda | linggao: also, moving the IPMI credential storage out of its current location and into a mroe secure location would help all drivers | 16:34 |
| anteaya | morning NobodyCam | 16:34 |
| anteaya | NobodyCam: how are you doing? | 16:35 |
| NobodyCam | good as can be.. :) thank you for asking, How are you :) | 16:35 |
| anteaya | :D | 16:35 |
| anteaya | good thanks | 16:35 |
| linggao | devananda, all the console support, should one bp enough or should I separate them into deifferent BPs? | 16:35 |
| anteaya | I slept so much after the summit, I was exhausted | 16:35 |
| *** martyntaylor has quit IRC | 16:36 | |
| devananda | linggao: one spec / bp should be fine, with several "work items" | 16:36 |
| devananda | linggao: eg, add common code to APIs; implement it for ipmitool; implement for ipminative; implement in python client; implement bindings in nova; -- and add tests at each phase | 16:37 |
| devananda | linggao: or something like that | 16:37 |
| *** BadCub has left #openstack-ironic | 16:38 | |
| linggao | devananda, cool, I like that. how about reset service processor (bmc)? Do you think it needs to get into Juno? | 16:38 |
| devananda | linggao: we're tight on review bandwidth. please file a spec for that -- i agree it's a necessary feature -- but i can't promise what we'll have time for, and that's lower priority than things which affect graduation | 16:39 |
| linggao | devananda, sure. I'll file a spec first. | 16:39 |
| linggao | thanks a lot. devananda. | 16:40 |
| devananda | linggao: np. thank you! | 16:40 |
| linggao | Hi anteaya, heard you have started a weekly meeting on 3rd party CI. | 16:41 |
| linggao | anreaya, when is it and what is the name of the IRC channel? | 16:41 |
| anteaya | https://wiki.openstack.org/wiki/Meetings/ThirdParty | 16:41 |
| anteaya | the irc channel is where ever you happen to be | 16:42 |
| anteaya | it is cross project so the meeting is the locus | 16:42 |
| anteaya | Mondays at 1800 utc in #openstack-meeting | 16:42 |
| *** romcheg has quit IRC | 16:43 | |
| linggao | anteaya, thanks for the info. | 16:43 |
| anteaya | np | 16:43 |
| anteaya | it would also be useful to attend or at least read the logs of the -infra, -tc and -qa meetings | 16:44 |
| linggao | ok. | 16:44 |
| anteaya | since the discussions there will have an impact on third party ci systems | 16:44 |
| anteaya | linggao: do you know how to find out about those meeting and access the logs? | 16:44 |
| anteaya | if no, i can show you | 16:44 |
| linggao | anteaya, no. please show me. | 16:45 |
| anteaya | okay | 16:45 |
| *** martyntaylor has joined #openstack-ironic | 16:45 | |
| lucasagomes | ok I'm done for the day, have a good night everybody! | 16:45 |
| anteaya | go to wiki.openstack.org > online meetings | 16:45 |
| anteaya | lucasagomes: good night | 16:45 |
| devananda | g'night lucasagomes | 16:45 |
| *** lucasagomes is now known as lucas-dinner | 16:45 | |
| anteaya | linggao: if you scroll down online meetings you will see project infrastructure (infra), technical committee (tc) and quality assurance (qa) | 16:46 |
| anteaya | linggao: can you find all of those? | 16:46 |
| *** zdiN0bot has joined #openstack-ironic | 16:46 | |
| anteaya | let me know if you can't | 16:46 |
| NobodyCam | have a good night lucas-dinner | 16:47 |
| *** lucas-dinner has quit IRC | 16:48 | |
| linggao | anteaya, I cannot find the quality assurance. | 16:48 |
| anteaya | kk | 16:49 |
| anteaya | https://wiki.openstack.org/wiki/Meetings/QATeamMeeting | 16:50 |
| anteaya | they are calling it QA | 16:50 |
| anteaya | sorry about that | 16:50 |
| linggao | anteaya, I found it. | 16:51 |
| anteaya | great | 16:51 |
| anteaya | so those are the meeting times and agendas | 16:51 |
| anteaya | and that is the list of all openstack meetings, so you should find any meeting you need to listed there | 16:51 |
| anteaya | next up, logs | 16:52 |
| anteaya | eavesdrop.openstack.org | 16:52 |
| linggao | ok | 16:52 |
| anteaya | the channel logs are under channel and meeting logs under meeting | 16:52 |
| anteaya | find the meeting log for yesterday's third party meeting | 16:52 |
| anteaya | then link me | 16:52 |
| linggao | I found the log yeaterday's third party meting log. | 16:54 |
| *** matty_dubs|lunch is now known as matty_dubs | 16:54 | |
| linggao | what do you mean by "link me" ? | 16:54 |
| *** harlowja_away is now known as harlowja_ | 16:55 | |
| anteaya | post the url for the log you found in channel | 16:55 |
| anteaya | so that I may click | 16:55 |
| linggao | http://eavesdrop.openstack.org/meetings/third_party/2014/third_party.2014-05-26-18.00.txt | 16:56 |
| anteaya | great | 16:56 |
| anteaya | good work | 16:56 |
| *** Alexei_9871 has quit IRC | 16:56 | |
| anteaya | now find the log for the same meeting in the channel logs | 16:56 |
| linggao | http://eavesdrop.openstack.org/irclogs/%23openstack-meeting/%23openstack-meeting.2014-05-26.log | 16:58 |
| linggao | this is the correct one? | 16:58 |
| anteaya | yes | 16:58 |
| anteaya | great work | 16:58 |
| anteaya | we are done | 16:58 |
| anteaya | you have now learned what you need to know to find the agenda for any meeting | 16:59 |
| anteaya | and the meeting log for any meeting | 16:59 |
| anteaya | and the channel log for any logged channel in openstack | 16:59 |
| anteaya | congratulations | 16:59 |
| linggao | anteaya, you are a great teacher. very patient, nice. | 16:59 |
| anteaya | :D | 16:59 |
| anteaya | thanks for your time | 16:59 |
| linggao | thank you! | 16:59 |
| anteaya | you're welcome | 17:00 |
| *** zdiN0bot has quit IRC | 17:00 | |
| linggao | ^_^ | 17:00 |
| anteaya | :D | 17:00 |
| *** igordcard has joined #openstack-ironic | 17:12 | |
| *** vinbs has quit IRC | 17:22 | |
| lifeless | devananda: this isn't about supporting LVM | 17:29 |
| lifeless | NobodyCam: hi | 17:29 |
| NobodyCam | lifeless: hi | 17:31 |
| *** romcheg has joined #openstack-ironic | 17:32 | |
| *** romcheg has quit IRC | 17:33 | |
| NobodyCam | lifeless: so how does switching to sfdisk sound. that should get rid of the label issue, then another patch to wide start and end of disk ? | 17:33 |
| NobodyCam | I'm looking to do what it takes to get things working for everyone | 17:33 |
| praefect | Hi guys, I'm trying to test ironic in devstack (linuxbridge), the nodes boot up (libvirt) but their DHCP requests never get answered, which is normal since their MAC never get added in the dhcp host file of the dnsmasq process running in dhcp namespace.. am I missing something? is this dhcp server only for instances or is it used also for PXE boot of the bm nodes? | 17:33 |
| lifeless | NobodyCam: lucas spent a fair chunk of time on that; the wipe seems like its a shorter path to get *working* again. | 17:35 |
| lifeless | NobodyCam: sounds like we need to get devananda's attention, going from backscroll | 17:35 |
| NobodyCam | ok so you read this morning chat | 17:36 |
| NobodyCam | :) | 17:36 |
| lifeless | highlights only | 17:36 |
| NobodyCam | devananda: ping | 17:36 |
| devananda | hi! | 17:36 |
| NobodyCam | :) hey hey | 17:36 |
| lifeless | praefect: the one DHCP server - the neutron dhcp server for the subnet - does PXE and normal address allocation | 17:37 |
| lifeless | devananda: so, I saw some concern about the disk start and end wiping strategy | 17:37 |
| lifeless | devananda: related to fuel wanting Ironic to do LVM ? | 17:38 |
| devananda | lifeless: in the scrollback, there was a lengty conversation before i came online. i responded to bits of it that i skimmed | 17:39 |
| lifeless | devananda: ok, so I fear you've misinterpreted | 17:39 |
| devananda | lifeless: wherein it seemed folks believed that the wipe-start-and-end was to erase LVM data | 17:39 |
| dtantsur|afk | rloo, addressed some of your comments here: https://review.openstack.org/#/c/92625/ please let me know what you think | 17:39 |
| devananda | which is not what I understood it to be -- and Iwas making clear my objection to LVM/MDADM | 17:39 |
| lifeless | devananda: the wipe start and end is to pretend to have a blank disk without writing 0's to the entire disk | 17:39 |
| devananda | lifeless: right -- I get that. there are partition tables and backup copies in those areas | 17:40 |
| lifeless | devananda: the amount that needs to be written to achieve that is the union of sizes of metadata schemes of all sorts | 17:40 |
| devananda | lifeless: i suspect that i'm not disagreeing with you at all | 17:40 |
| lifeless | not just the ones we support writing | 17:40 |
| lifeless | devananda: cool, I'm just seeking clear consensus | 17:40 |
| lifeless | devananda: so LVM gets mentioned, because its such a scheme we need to be sure to wipe, or some poor schmuck turning up with a previously used disk gets the short end of the stick | 17:41 |
| *** eghobo has quit IRC | 17:41 | |
| dtantsur|afk | devananda, hi! Could you have a look at https://review.openstack.org/#/c/95573/ ? It's fix for urgent problem with NodeLocked and is somewhat controversial. I've put my points in commit message and comments. | 17:41 |
| *** eghobo has joined #openstack-ironic | 17:42 | |
| devananda | dtantsur|afk: yea, been mulling that over while taking care of other things | 17:42 |
| dtantsur|afk | thanks! | 17:42 |
| *** dwalleck has joined #openstack-ironic | 17:42 | |
| NobodyCam | I'm happy to work the fix thru to the end I just don't want to spin our wheels for another week | 17:43 |
| devananda | lifeless: ack | 17:43 |
| lifeless | NobodyCam: that sounds like devananda and I are in consensus - | 17:43 |
| praefect | lifeless: thanks so I have to see the bm MAC in there somehow I guess… | 17:43 |
| devananda | NobodyCam: can you summarize all the objections at this point? I've lost track ... it seems like there is bikeshedding on exactly how many bytes to erase, when that's not the point | 17:43 |
| lifeless | devananda: there was, I think we cleared that up yesterday | 17:43 |
| lifeless | devananda: when I pointed out it wasnt' about the partition types we understand | 17:44 |
| lifeless | devananda: there is a functional problem that needs fixing in the code (touching the partition table at all on a rebuild-preserve-ephemeral) but that issue pre-exists, however we need to fix it here | 17:45 |
| devananda | dtantsur|afk: my original objection to using 409 is that it's ultimately not a client-side-error at all | 17:45 |
| NobodyCam | we started with the wipe at 4k then droped it to 1k (which corrupts the gpt table) but does not wipe it. | 17:45 |
| NobodyCam | looking at the gpt specs we need to wipe 18k (or 16 cann't recall right now) | 17:45 |
| dtantsur|afk | devananda, my point here is that it's not an error at all :) neither client not server. But at the end, server can't do anything about it, so it's responsibility of client - thus 4xx | 17:46 |
| devananda | dtantsur|afk: it's a server-side error in as much as we engineered it poorly and should never be exposing this situation at all | 17:46 |
| devananda | dtantsur|afk: api service should ACK that it got the request and stick it on the queue. done. hence my spec to fix this :) | 17:46 |
| dtantsur|afk | dtantsur|afk, if it is server-side error, nothing is wrong in reporting it with LOG.error, right? ;) And while I also prefer async approach generally, we have what we have for now | 17:47 |
| *** Lingo has joined #openstack-ironic | 17:47 | |
| dtantsur|afk | dtantsur|afk, reporting 5xx as error and failing test seems the right thing to do for me | 17:47 |
| dtantsur|afk | oh,why am I talking to myself? | 17:48 |
| devananda | NobodyCam: that wipe only needs to happen when (re)deploying though, I presume, and not on a rebuild-preserve-ephemeral | 17:48 |
| devananda | devananda: i dont know :) | 17:48 |
| dtantsur|afk | :D | 17:48 |
| rloo | dtantsur|afk: thx. | 17:48 |
| lifeless | devananda: correct (to NobodyCam ) - and I asked for that in my review | 17:48 |
| dtantsur|afk | devananda, on the other side with async approach, imagine we got 2 deploy requests for the same node | 17:48 |
| NobodyCam | devananda: labels need to be removed to use parted | 17:48 |
| lifeless | NobodyCam: its a more general problem, the parted stuff is just a sympotom | 17:49 |
| dtantsur|afk | devananda, I doubt we should make client wait before telling him that node is in incorrect state | 17:49 |
| NobodyCam | ack I was working on a patch that replaced parted with sfdisk | 17:49 |
| devananda | dtantsur|afk: sure. so I know of two approaches that solve taht | 17:49 |
| lifeless | NobodyCam: I think thats premature | 17:49 |
| *** dtantsur|afk is now known as dtantsur | 17:49 | |
| lifeless | NobodyCam: we may want to, but the issues we had with it were troubling | 17:50 |
| NobodyCam | ok | 17:50 |
| devananda | dtantsur|afk: 1. every request generates a request token. user can track that. (alternate: use callback to POST result back to user) | 17:50 |
| lifeless | NobodyCam: and making the disk look 'blank' on a new deploy is clearly the right thing to do | 17:50 |
| devananda | dtantsur: 2. certain API actions, like deploy, use an intent-lock. so concurrent requests will cause all-but-one to fail on API side, not waiting on RPC bus | 17:50 |
| dtantsur | devananda, I don't like that (1) will require to store tokens in DB to make API instances stateless | 17:51 |
| dtantsur | devananda, (2) sounds find. So periodic tasks don't fail things - should work! | 17:52 |
| *** zdiN0bot has joined #openstack-ironic | 17:52 | |
| dtantsur | (though in my previous experience, we preferred to let client decide about retrying/not retrying, but it was somewhat different application) | 17:53 |
| *** dwalleck has quit IRC | 17:53 | |
| dtantsur | devananda, returning to the current problem: whatever solution we find now, it should be temporary, right? What do you think about 409 as a temporary solution? | 17:53 |
| devananda | dtantsur: have you reviewed https://review.openstack.org/#/c/94923/5/specs/juno/fully-async-api.rst ? | 17:54 |
| *** dwalleck has joined #openstack-ironic | 17:54 | |
| dtantsur | devananda, I actually read it and left come comments. After this chat btw I understand your motivation better, so it may be useful to put some points there | 17:54 |
| dtantsur | e.g. alternatives on async approach | 17:55 |
| dtantsur | devananda, actually, I hope to get more ideas tomorrow, as I got seriously tired today :( | 17:58 |
| *** dwalleck has quit IRC | 17:58 | |
| dtantsur | So wishing you all good night :) | 17:58 |
| rloo | dtantsur: let me know if you have time to chat about 92625. | 17:58 |
| dtantsur | rloo, well, if it's not too long :) | 17:59 |
| rloo | dtantsur: we can chat tomorrow then. Or I will just add more comments. | 17:59 |
| dtantsur | rloo, can I just answer something for you? | 17:59 |
| devananda | dtantsur: i'll keep thinking on this one, and try to find another fix for the gate today. good night :) | 17:59 |
| rloo | dtantsur: wrt total_size < free or <= free. I don't really care, it is just that your if is total_size > free, not total_size >= free :-) | 18:00 |
| *** rloo has quit IRC | 18:00 | |
| *** rloo has joined #openstack-ironic | 18:00 | |
| NobodyCam | devananda: lifeless: so stick with wiping start and end of disk, rework to not repartition on rebuilds with preserve_ephemeral set? | 18:01 |
| *** dwalleck has joined #openstack-ironic | 18:02 | |
| dtantsur | rloo, oh yes, didn't notice :) I actually see no serious difference, but I can fix it to some consistent way | 18:03 |
| rloo | dtantsur: ok thx. Sorry about that, just made me wonder why it was off by 1... | 18:04 |
| dtantsur | rloo, after speaking to you I realized we need to somehow make room also for converting... But that's not for this patch, I guess | 18:04 |
| rloo | dtantsur: don't add more functionality ;) | 18:05 |
| dtantsur | yeah, sometimes enough is enough :) | 18:05 |
| rloo | dtantsur: i added comments. the test concerns me though. seems odd that the syntax is different, [] vs (). | 18:05 |
| lifeless | NobodyCam: I think if you do precisely what my review asks for, you'll be clear with deva and I | 18:05 |
| lifeless | NobodyCam: which was a little more than your last sentence, but yes. | 18:05 |
| dtantsur | rloo, it's just testing 2 functions that call the same function in a bit different way | 18:05 |
| NobodyCam | ack | 18:06 |
| rloo | lifeless, NobodyCam: but the intent is to address more that the volume label thing right? | 18:06 |
| *** dwalleck has quit IRC | 18:06 | |
| NobodyCam | brb | 18:06 |
| *** dwalleck has joined #openstack-ironic | 18:06 | |
| rloo | dtantsur: looking... | 18:06 |
| dtantsur | rloo, it does not matter for sequence unpacking (that's what I do in fetch_images), but it does matter for equality comparison | 18:07 |
| rloo | dtantsur: so you're saying that pxe._fetch_images(), the images_info argument can be different? | 18:07 |
| dtantsur | images_info = sequence(sequence(uuid, path), sequence(uuid, path)), where sequence can be whatever: tuple, list... | 18:08 |
| dtantsur | the only place where it matters is equality comparison in tests | 18:08 |
| rloo | dtantsur: would you mind adding a docstring about images_info then? Or do you think it doesn't need a docstring? | 18:09 |
| rloo | dtantsur: maybe it is just me. I assumed it was a tuple. | 18:10 |
| lifeless | rloo: sorry, don't quite parse that | 18:10 |
| dtantsur | rloo, I wish it was tuple everywhere, it is more correct; but values in pxe_info are lists | 18:11 |
| rloo | lifeless. Well, NobodyCam's commit msg is misleading i think. | 18:11 |
| dtantsur | rloo, I can add a docstring if you want it confusing | 18:11 |
| rloo | lifeless: "Workaround for conflicting volume labels". | 18:12 |
| dtantsur | I can even start using lists everywhere for the sake of consistency :) | 18:12 |
| lifeless | rloo: oh certainly, that should be fixed too | 18:12 |
| rloo | lifeless, good. NobodyCam: ^^ | 18:12 |
| rloo | dtantsur: whichever you think is easier for people to understand/code to maintain/whatever. | 18:13 |
| NobodyCam | will change... but that is how the patch started life | 18:13 |
| NobodyCam | :-p | 18:13 |
| rloo | NobodyCam: did you create a monster? | 18:13 |
| rloo | NobodyCam: err, I mean, a butterfly? | 18:14 |
| NobodyCam | lol seems so :-p (/me has seen the butterfly effect :-p) | 18:14 |
| dtantsur | rloo, problem is that it's ok from my point of view :) line 387: image_info[label] = [None, None] <-- that's what I would not do, this should be a tuple by it's sense | 18:15 |
| dtantsur | rloo, what if I change this line to a tuple? That will make code consistent and more correct | 18:16 |
| rloo | dtantsur: that makes sense with me. At first glance anyway; I'm not up-to-date on that code. Is that the only place where it is a list? | 18:17 |
| dtantsur | rloo, at first glance yes. I'll check better when I'll be changing it | 18:18 |
| rloo | dtantsur: thx. if there are more places, then I'd say clean it up in another patch. | 18:18 |
| dtantsur | rloo, ack, will fix, thank you for helping clarify it :) | 18:18 |
| rloo | dtantsur: thx. | 18:19 |
| dtantsur | that's all for tomorrow, g'night now | 18:19 |
| rloo | dtantsur: good night! | 18:19 |
| *** dtantsur is now known as dtantsur|afk | 18:19 | |
| NobodyCam | night dtantsur | 18:19 |
| *** krtaylor has quit IRC | 18:22 | |
| *** sysexit has quit IRC | 18:23 | |
| *** romcheg has joined #openstack-ironic | 18:27 | |
| *** romcheg has quit IRC | 18:27 | |
| *** sysexit has joined #openstack-ironic | 18:49 | |
| *** rloo has quit IRC | 18:49 | |
| *** rloo has joined #openstack-ironic | 18:49 | |
| *** rloo has quit IRC | 18:50 | |
| *** zdiN0bot has quit IRC | 18:50 | |
| *** rloo has joined #openstack-ironic | 18:50 | |
| *** rloo has quit IRC | 18:51 | |
| *** rloo has joined #openstack-ironic | 18:51 | |
| *** max_lobur has quit IRC | 18:56 | |
| *** dwalleck has quit IRC | 19:01 | |
| *** derekh_ has joined #openstack-ironic | 19:02 | |
| *** rloo has quit IRC | 19:03 | |
| *** rloo has joined #openstack-ironic | 19:03 | |
| *** Penick has joined #openstack-ironic | 19:04 | |
| *** Lingo is now known as BadCub | 19:05 | |
| *** sysexit has quit IRC | 19:09 | |
| *** rloo has quit IRC | 19:09 | |
| *** rloo has joined #openstack-ironic | 19:09 | |
| *** dwalleck has joined #openstack-ironic | 19:15 | |
| *** ifarkas has quit IRC | 19:27 | |
| *** sysexit has joined #openstack-ironic | 19:37 | |
| linggao | devanana, I am working on the console log specs. Just want to confirm with you that nova-baremtal did not implement the get-console-output, correct? | 19:40 |
| devananda | linggao: afaik, correct. nova-bm only has support for shellinaboxd | 19:41 |
| linggao | which is the interactive console. | 19:42 |
| linggao | In order to use console log, we must save the console logs. | 19:42 |
| linggao | The it comes when to save it and where to save it. Any suggestions? | 19:43 |
| linggao | and how to age it out. | 19:44 |
| devananda | linggao: nope. i haven't researched that yet. i would start by looking at what nova does for VM's console logs and see if we can follow that precedent | 19:46 |
| matty_dubs | That sounds like it could get icky. :-\ | 19:47 |
| matty_dubs | (Though I agree with the approach.) | 19:47 |
| linggao | ok, I'll start from that route first. | 19:47 |
| linggao | jbjhonso has a python lib called confluent, which is a console server that stores the console logs. | 19:49 |
| linggao | The console can be accessed through command line as welll as http. | 19:51 |
| *** rloo has quit IRC | 19:57 | |
| *** rloo has joined #openstack-ironic | 19:58 | |
| devananda | python-ironicclient is now gating symmetrically with ironic | 19:59 |
| openstackgerrit | Martin Geisler proposed a change to openstack/python-ironicclient: Use Emacs-friendly file variable to set file encoding https://review.openstack.org/95882 | 19:59 |
| devananda | meaning tempest tests should run on the client too ,now | 19:59 |
| *** dwalleck has quit IRC | 20:00 | |
| NobodyCam | brb | 20:01 |
| rloo | devananda: yay for ironicclient! | 20:02 |
| *** zdiN0bot has joined #openstack-ironic | 20:06 | |
| *** krtaylor has joined #openstack-ironic | 20:07 | |
| *** BadCub01_ has joined #openstack-ironic | 20:08 | |
| *** BadCub has quit IRC | 20:08 | |
| *** boris-42 has quit IRC | 20:21 | |
| *** boris-42 has joined #openstack-ironic | 20:21 | |
| linggao | devananda. with VM hypervisors, there is typically a serial console or a paravirtualized log for each virtual machine. | 20:22 |
| linggao | It is exposed on the host as pty or a plian file. | 20:23 |
| linggao | Nova libvirt driver just uses it to get console output. | 20:23 |
| linggao | With baremetal, we need to deside when and where to store them. xCAT used conserver, now is using confluent. | 20:25 |
| linggao | devananda, So we have 2 options. 1 use third party console server to log and access the console. 2. Home grow our own console server in Ironic, we can use the ironic-conductor or a new daemon as the console server, spawn a process to keep saving the serial console to a file when the node is being powered on. The file can be flushed with next power on action. | 20:34 |
| linggao | The console file can be saved on the host where the conductor is to begin with. Later we can add options to save it to user-supplied storeage devices. | 20:37 |
| *** jbjohnso has quit IRC | 20:39 | |
| devananda | linggao: storing console log on conductor is fragile -- what happens when the conductor ring rebalances? | 20:40 |
| linggao | hmm, | 20:41 |
| *** BadCub has joined #openstack-ironic | 20:41 | |
| linggao | should the console log be moved to the new host when the conductor ring rebalances? | 20:42 |
| *** BadCub has quit IRC | 20:42 | |
| linggao | I do not know the details of the rebalance. do we have control over it? | 20:43 |
| *** romcheg has joined #openstack-ironic | 20:43 | |
| linggao | Or, start the new console log on the new host. you just loose the old console after rebalance. | 20:44 |
| devananda | rebalance can occur eg. when a new conductor joins the ring, or one fails and is removed automaticaly | 20:44 |
| devananda | so we can't guarantee to preserve local files | 20:44 |
| devananda | storing in database is one option, though not necessarily scalable | 20:45 |
| devananda | start new log and lose old content is an option, but not very appealing to operators | 20:45 |
| matty_dubs | Oh, I see -- so Nova can just dump to local disk, since if the host goes, the VM goes too. But we can't, since we could lose the conductor (or just rebalance) without the BM node going down? | 20:45 |
| devananda | right | 20:45 |
| linggao | loosing console should not be a big deal, user usually want to log at the console when the deployment fails. | 20:46 |
| linggao | they can just restart the node to get the new console logs. :-) | 20:47 |
| linggao | s/log/look/ | 20:47 |
| linggao | another way is find a cinder storage for the console logs. | 20:49 |
| linggao | that can be an option. | 20:50 |
| *** jdob has quit IRC | 20:50 | |
| *** zdiN0bot has quit IRC | 20:50 | |
| matty_dubs | It seems like "I'd like to persistently store and write to this log" shouldn't be an Ironic problem. | 20:51 |
| matty_dubs | Though obviously, right now, it is. | 20:51 |
| devananda | matty_dubs: it's not a user-writable log | 20:52 |
| matty_dubs | Oh, "I" was Ironic, vs. an admin. | 20:52 |
| devananda | matty_dubs: most hardware's BMC exposes a small buffer containing the last X kb of kernel log | 20:52 |
| devananda | ah | 20:52 |
| devananda | nvm | 20:53 |
| linggao | nvm? I am googling... | 20:54 |
| matty_dubs | Better-stated, it seems like 'central logging' ought not to be an Ironic-specific task. | 20:55 |
| matty_dubs | nvm == nevermind | 20:55 |
| matty_dubs | Though not that I'm suggesting we should throw our hands up and say "Not our problem!" I just wonder if other projects are having this same issue. | 20:56 |
| linggao | lol, I found node version manager. :) | 20:56 |
| matty_dubs | <strike>We could write a log store in Node.js!</strike> | 20:57 |
| *** praefect has quit IRC | 20:58 | |
| matty_dubs | devananda: I don't know much about BMC buffers, beyond basic hardware errors showing up there. But your comment about buffers makes me wonder if "Don't log stuff ourselves, but read from the BMC log if something goes wrong" is a viable option? | 20:59 |
| *** romcheg has quit IRC | 21:00 | |
| linggao | it has hardware event logs, like node powered on or off, firmare updated etc. I do no think it has sol console log. | 21:02 |
| matty_dubs | Ah, that might explain why my Googling wasn't turning up anything | 21:02 |
| *** Penick has quit IRC | 21:02 | |
| linggao | :) | 21:02 |
| *** Penick has joined #openstack-ironic | 21:04 | |
| linggao | basically the console server will enalbe sol on BMC and saves any sol output real time to file in order to support console log. | 21:04 |
| matty_dubs | Ah, okay. I think I misunderstood. | 21:05 |
| matty_dubs | Wishful thinking on my part. ;) | 21:05 |
| linggao | It could cause a lot of traffic if a lot of nodes are being deployed at the same time. That's why the conserverd has an option that only saves the console for the ones you are looking at interactively. | 21:06 |
| devananda | linggao: i was under the (perhaps incorrect) assumption that BMCs often maintain a serial console buffer as well, meaning one doesn't necessarily need continuous SOL connection to capture the serial output | 21:07 |
| matty_dubs | AFAICT that's not standard IPMI, at least, but I'm going only by what Google is telling me. | 21:08 |
| devananda | linggao: the intention of this feature is to aid in debugging failed deployments. so the operator would need the console to have been logged once the deploy started, but /before/ realizing that the deploy failed | 21:09 |
| linggao | I asked jbjohnso if BMC stores the console itself, he said no. | 21:09 |
| devananda | linggao: ok, thanks | 21:09 |
| linggao | devananda, true. | 21:11 |
| devananda | there is unfortunately no efficient way to store log files in MySQL | 21:11 |
| devananda | storing as local files on the conductor is certainly the easiest solution | 21:12 |
| linggao | right. we can begin with this. | 21:14 |
| *** romcheg has joined #openstack-ironic | 21:14 | |
| linggao | devananda, do you prefer use a 3rd party console server or create our own? | 21:15 |
| devananda | linggao: not entirely sure what you mean | 21:16 |
| *** dwalleck has joined #openstack-ironic | 21:16 | |
| linggao | I mean use conserver or confluent, or implement the same functions in Ironic. | 21:16 |
| Shrews | devananda: question re: https://review.openstack.org/95213 ... since that changes behavior *after* a release, how is that dealt with? seems like there should at least be something documenting the behavior change somewhere | 21:17 |
| linggao | basically they keep saving the consoles output in console log and making sure the connection is alive all the time. | 21:17 |
| devananda | linggao: can you paste links to those project's home pages? | 21:19 |
| linggao | http://www.conserver.com/ | 21:19 |
| devananda | Shrews: like a DocImpact flag, or a bug, or a change to docs/...? | 21:20 |
| matty_dubs | I've got to head out; have a great day/night linggao and devananda (et al.)! | 21:21 |
| devananda | matty_dubs: you too! | 21:21 |
| linggao | confluent is something jbjohnso is developing (console) for xCAT and openstack. I only know where is the source is. | 21:22 |
| linggao | http://sourceforge.net/p/xcat/confluent/ci/master/tree/ | 21:22 |
| Shrews | devananda: Something like that. If someone is upgrading from I to J, this change would be altering their existing setup, so they'd need to be aware of the new config to reproduce what they had in I. | 21:22 |
| *** matty_dubs is now known as matty_dubs|gone | 21:22 | |
| devananda | Shrews: like upgrade notes :) | 21:23 |
| Shrews | devananda: if that is even a concern at this point... thus my question :) | 21:23 |
| linggao | devananda, but I know jbjohnso always have perfoamnce in mind when he develops code. | 21:23 |
| devananda | Shrews: let's assume there will be. then in the worst case, we didn't make it worse :) | 21:23 |
| linggao | matty_dubs, good night, | 21:24 |
| linggao | devananda, I also needs to go. talk to you tomorrow. | 21:24 |
| linggao | s/needs/need/ :) | 21:25 |
| *** mrda_away is now known as mrda | 21:28 | |
| mrda | Morning Ironic! | 21:28 |
| Shrews | mrda: g'morning | 21:29 |
| devananda | mornin! | 21:29 |
| mrda | \o | 21:29 |
| *** linggao has quit IRC | 21:29 | |
| -openstackstatus- NOTICE: Zuul is offline due to an operational issue; ETA 2200 UTC. | 21:32 | |
| *** ChanServ changes topic to "Zuul is offline due to an operational issue; ETA 2200 UTC." | 21:32 | |
| *** Penick has quit IRC | 21:38 | |
| *** davidlenwell_ has joined #openstack-ironic | 21:40 | |
| *** davidlenwell_ has quit IRC | 21:44 | |
| *** davidlenwell has quit IRC | 21:45 | |
| *** kaizer has quit IRC | 21:46 | |
| *** davidlenwell has joined #openstack-ironic | 21:47 | |
| NobodyCam | morning mrda | 21:50 |
| *** romcheg has quit IRC | 21:51 | |
| *** harlowja_ is now known as harlowja_away | 21:57 | |
| *** harlowja_away is now known as harlowja_ | 21:57 | |
| NobodyCam | starts tox on a rebuild and steps afk | 22:01 |
| *** shakamunyi has quit IRC | 22:03 | |
| *** dwalleck has quit IRC | 22:08 | |
| *** romcheg has joined #openstack-ironic | 22:08 | |
| -openstackstatus- NOTICE: Zuul is started and processing changes that were in the queue when it was stopped. Changes uploaded or approved since then will need to be re-approved or rechecked. | 22:09 | |
| *** ChanServ changes topic to "OpenStack Bare Metal Provisioning | Docs: http://docs.openstack.org/developer/ironic/ | Bugs: https://bugs.launchpad.net/ironic | Status: https://etherpad.openstack.org/p/IronicWhiteBoard" | 22:09 | |
| *** shakamunyi has joined #openstack-ironic | 22:10 | |
| adam_g | devananda, wondering if you had any input on https://bugs.launchpad.net/ironic/+bug/1321504 | 22:23 |
| devananda | adam_g: no - i haven't the hardware to run that test on, but very excited that you are! | 22:24 |
| devananda | adam_g: initial guess would be aroudn 3 areas: network/io contention in copying the image. you seem to be working on that | 22:25 |
| devananda | and there are a few other patches touching it too | 22:25 |
| devananda | or - scheduler picking same node repeatedly and triggering races, but that would show up in logs | 22:25 |
| devananda | or - some unknown contention within ir-cond or n-cpu | 22:26 |
| adam_g | devananda, yeah--i was wondering what you thought about limiting the # of concurrent deployments to a pool of workers | 22:26 |
| adam_g | that patch fixes the fail rate but increases deployment time considerably | 22:27 |
| *** jgrimm has quit IRC | 22:27 | |
| devananda | adam_g: yea, i expect it would increase deploy time | 22:28 |
| devananda | adam_g: i'd like to see a pool of deploy workers taht can scale based on load | 22:28 |
| devananda | or be sized by the operator according to their network / io cap | 22:29 |
| devananda | or something | 22:29 |
| adam_g | devananda, interesting. | 22:29 |
| adam_g | yeah i was just thinking a config flag that an admin can tweak accordingly | 22:29 |
| devananda | though IIRC, lifeless disagreed with that approach | 22:29 |
| devananda | clearly, allowing an unlimited # of concurrent deploys is going to expose serious failure modes where the network gets saturated -- even after we move to an agent-based-download | 22:30 |
| devananda | the saturation will move off of the conductor and onto the image source -- which is much easier to scale (add a proxy, or more glance backends, etc) but the network itself is not infinite | 22:31 |
| lifeless | hi | 22:31 |
| devananda | adam_g: https://review.openstack.org/#/c/95213/ | 22:31 |
| lifeless | so my main concern is that we don't pessimise first time-to-live by trading throughput for concurrency | 22:31 |
| lifeless | greg says that the serialised version is currently a pessimisation | 22:32 |
| lifeless | we haven't looked into why | 22:32 |
| adam_g | devananda, yeah, i saw that. funny that came up the same time i hit this | 22:32 |
| *** sysexit has quit IRC | 22:32 | |
| lifeless | the testbed we're working on has 10Gbps networking, and lots of memory | 22:32 |
| devananda | adam_g: i suspect the RH folks hit the same issue, and thus wrote that | 22:32 |
| lifeless | so we should be able to write entirely to RAM on the recipient machines in short order | 22:32 |
| adam_g | lifeless, receipient machine in this case is iscsi target / ironic node? | 22:33 |
| lifeless | we already have https://bugs.launchpad.net/ironic/+bug/1199665 | 22:33 |
| lifeless | adam_g: yes | 22:33 |
| lifeless | adam_g: I suspect the reason the serialised version is slower right now is that dd calls fsync at the end of the copy | 22:34 |
| lifeless | and that will block until there are no dirty pages on the device | 22:34 |
| adam_g | oflag=direct | 22:34 |
| lifeless | adam_g: yes | 22:34 |
| adam_g | what about the targets? | 22:34 |
| lifeless | adam_g: that means we don't keep a pointless copy in the page cache on the local system | 22:34 |
| adam_g | are they configured for async? | 22:34 |
| lifeless | adam_g: good question | 22:35 |
| devananda | or are the iscsi targets set up for directio? | 22:35 |
| adam_g | devananda, yeah, im curious | 22:35 |
| devananda | that'd prevent the optimizations you're thinking of | 22:36 |
| devananda | and i remember us doing that over a year ago | 22:36 |
| lifeless | I do like the serialisation thing in general | 22:36 |
| lifeless | but | 22:36 |
| lifeless | some network topologies (e.g. moonshot) will want to muliplex from the conductor | 22:37 |
| lifeless | (e.g. 10Gbps or more on the conductor, 1Gbps per recipient) | 22:37 |
| lifeless | so having a knob is likely necessary (though it would be great to make it automatic in some fashion eventually | 22:37 |
| devananda | sure | 22:39 |
| *** romcheg has quit IRC | 22:39 | |
| devananda | i don't see anything in DIB or ironic that is specifying direct IO for tgtadm | 22:42 |
| devananda | eg, http://manpages.ubuntu.com/manpages/trusty/man8/tgtadm.8.html#contenttoc6 | 22:42 |
| *** Penick has joined #openstack-ironic | 22:43 | |
| lifeless | thats good | 22:44 |
| lifeless | sync we want the pagecache on the target | 22:44 |
| lifeless | since | 22:44 |
| adam_g | -e aio? | 22:47 |
| devananda | adam_g: i was just thinking the same thing | 22:47 |
| devananda | worth a try | 22:47 |
| adam_g | i find it hard to believe tgt would default to writing to the page cache | 22:47 |
| *** derekh_ has quit IRC | 22:51 | |
| devananda | ok - i must go afk and find food. will bbiah | 22:52 |
| *** zdiN0bot has joined #openstack-ironic | 22:57 | |
| *** zdiN0bot has quit IRC | 23:00 | |
| *** zdiN0bot has joined #openstack-ironic | 23:00 | |
| *** rloo has quit IRC | 23:04 | |
| *** rloo has joined #openstack-ironic | 23:04 | |
| *** zdiN0bot has quit IRC | 23:28 | |
| *** igordcard has quit IRC | 23:29 | |
| *** Penick has quit IRC | 23:30 | |
| *** rloo has quit IRC | 23:44 | |
| *** rloo has joined #openstack-ironic | 23:44 | |
| *** eguz has joined #openstack-ironic | 23:51 | |
| *** zdiN0bot has joined #openstack-ironic | 23:52 | |
| *** zdiN0bot1 has joined #openstack-ironic | 23:55 | |
| *** eghobo has quit IRC | 23:55 | |
| *** zdiN0bot has quit IRC | 23:57 | |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Allow configdrive partition to be precreated https://review.openstack.org/95969 | 23:58 |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Allow configdrive partition to be precreated https://review.openstack.org/95969 | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!