| *** zdiN0bot has quit IRC | 00:03 | |
| *** zdiN0bot has joined #openstack-ironic | 00:05 | |
| *** zdiN0bot has quit IRC | 00:22 | |
| *** ilives has joined #openstack-ironic | 00:22 | |
| *** zdiN0bot has joined #openstack-ironic | 00:26 | |
| *** ilives has quit IRC | 00:27 | |
| *** ilives has joined #openstack-ironic | 00:28 | |
| openstackgerrit | A change was merged to openstack/ironic: Add DiskPartitioner https://review.openstack.org/83396 | 00:30 |
|---|---|---|
| *** matsuhashi has joined #openstack-ironic | 00:31 | |
| *** blamar has quit IRC | 00:34 | |
| *** zdiN0bot1 has joined #openstack-ironic | 00:52 | |
| *** zdiN0bot has quit IRC | 00:54 | |
| *** blamar has joined #openstack-ironic | 00:59 | |
| *** newell_ has quit IRC | 01:03 | |
| *** zdiN0bot1 has quit IRC | 01:06 | |
| *** eguz has quit IRC | 01:17 | |
| *** todd_dsm has joined #openstack-ironic | 01:32 | |
| *** epim has quit IRC | 01:32 | |
| *** nosnos has joined #openstack-ironic | 01:33 | |
| *** rloo has quit IRC | 01:41 | |
| *** jgrimm has joined #openstack-ironic | 01:52 | |
| *** ilives has quit IRC | 02:35 | |
| *** ilives has joined #openstack-ironic | 02:36 | |
| *** lnxnut_ has joined #openstack-ironic | 02:51 | |
| *** lnxnut has quit IRC | 02:53 | |
| *** matsuhashi has quit IRC | 03:14 | |
| *** nosnos has quit IRC | 03:27 | |
| *** lnxnut_ has quit IRC | 03:33 | |
| *** todd_dsm has quit IRC | 03:50 | |
| *** todd_dsm has joined #openstack-ironic | 03:51 | |
| *** matsuhashi has joined #openstack-ironic | 04:20 | |
| *** nosnos has joined #openstack-ironic | 04:23 | |
| *** vkozhukalov has joined #openstack-ironic | 04:32 | |
| *** eghobo has joined #openstack-ironic | 04:32 | |
| *** todd_dsm has quit IRC | 04:34 | |
| *** harlowja is now known as harlowja_away | 04:47 | |
| *** ilives has quit IRC | 04:49 | |
| *** todd_dsm has joined #openstack-ironic | 04:49 | |
| *** ilives has joined #openstack-ironic | 04:50 | |
| *** todd_dsm has quit IRC | 05:00 | |
| *** eghobo has quit IRC | 05:00 | |
| *** eghobo has joined #openstack-ironic | 05:00 | |
| *** todd_dsm has joined #openstack-ironic | 05:00 | |
| *** boris-42 has quit IRC | 05:24 | |
| *** boris-42 has joined #openstack-ironic | 05:43 | |
| *** vkozhukalov has quit IRC | 06:03 | |
| *** eghobo has quit IRC | 06:11 | |
| *** pradipta_away is now known as pradipta | 06:14 | |
| *** ifarkas has joined #openstack-ironic | 06:55 | |
| *** ekarlso has quit IRC | 07:12 | |
| *** datajerk has quit IRC | 07:17 | |
| *** ilives has quit IRC | 07:29 | |
| *** ilives has joined #openstack-ironic | 07:29 | |
| *** datajerk has joined #openstack-ironic | 07:31 | |
| *** ifarkas has quit IRC | 07:32 | |
| *** mrda is now known as mrda_away | 07:34 | |
| *** lsmola has quit IRC | 07:34 | |
| *** athomas has joined #openstack-ironic | 07:40 | |
| *** ndipanov has quit IRC | 07:43 | |
| *** ifarkas has joined #openstack-ironic | 07:44 | |
| *** max_lobur has joined #openstack-ironic | 07:54 | |
| *** martyntaylor has quit IRC | 07:57 | |
| *** ndipanov has joined #openstack-ironic | 07:58 | |
| *** vkozhukalov has joined #openstack-ironic | 07:58 | |
| *** lsmola has joined #openstack-ironic | 07:58 | |
| *** martyntaylor has joined #openstack-ironic | 08:09 | |
| *** lucasagomes has joined #openstack-ironic | 08:11 | |
| *** jistr has joined #openstack-ironic | 08:13 | |
| *** martyntaylor has quit IRC | 08:16 | |
| dtantsur | Morning Ironic | 08:21 |
| *** foexle has joined #openstack-ironic | 08:24 | |
| *** dshulyak has quit IRC | 08:26 | |
| *** martyntaylor has joined #openstack-ironic | 08:31 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 08:34 | |
| *** pradipta is now known as pradipta_away | 08:43 | |
| *** derekh has joined #openstack-ironic | 08:52 | |
| yuriyz | morning dtantsur | 09:02 |
| Mikhail_D_ltp | Morning all! :) | 09:02 |
| *** romcheg has quit IRC | 09:03 | |
| *** romcheg has joined #openstack-ironic | 09:03 | |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk utils https://review.openstack.org/86163 | 09:13 |
| Haomeng | morning all:) | 09:17 |
| *** Mikhail_D_ltp has left #openstack-ironic | 09:18 | |
| dtantsur | morning Haomeng, yuriyz, Mikhail_D_wk :) | 09:29 |
| lucasagomes | morning dtantsur Haomeng yuriyz Mikhail_D_wk | 09:33 |
| dtantsur | lucasagomes, morning | 09:33 |
| romcheg | Morning dtantsur lucasagomes and everyone else | 09:44 |
| dtantsur | romcheg, morning | 09:45 |
| lucasagomes | :) romcheg morning | 09:45 |
| * dtantsur 's job today is too greet everyone :) | 09:46 | |
| romcheg | dtantsur, :-P | 10:00 |
| *** dshulyak has joined #openstack-ironic | 10:01 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 10:27 | |
| *** lnxnut has joined #openstack-ironic | 10:44 | |
| *** Mikhail_D_ltp has quit IRC | 10:54 | |
| Haomeng | dtantsur: morning:) | 11:03 |
| *** dtantsur is now known as dtantsur|launch | 11:08 | |
| *** killer_prince has quit IRC | 11:13 | |
| *** lazy_prince has joined #openstack-ironic | 11:32 | |
| *** lazy_prince is now known as killer_prince | 11:32 | |
| *** lucasagomes is now known as lucas-hungry | 11:53 | |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk utils https://review.openstack.org/86163 | 12:10 |
| *** jdob has joined #openstack-ironic | 12:11 | |
| *** lucas-hungry is now known as lucasagomes | 12:29 | |
| *** mgagne has quit IRC | 12:31 | |
| *** mgagne has joined #openstack-ironic | 12:33 | |
| *** mgagne is now known as Guest78217 | 12:33 | |
| romcheg | lucasagomes: I'm looking at https://review.openstack.org/#/c/83572/ | 12:39 |
| lucasagomes | romcheg, right | 12:39 |
| lucasagomes | romcheg, that superceed the whitelist one | 12:39 |
| romcheg | right | 12:40 |
| lucasagomes | romcheg, it was created before the ff... I'll review it too | 12:40 |
| romcheg | lucasagomes: test_ports.py contains a lot of changes not related to the patch | 12:40 |
| lucasagomes | ew | 12:40 |
| romcheg | should we let it go or ask to separate them | 12:40 |
| lucasagomes | hmmm, the way I usually think about it is... if ur touching a specific method and u want to do a small cleanup/change that is not really related to the patch fine | 12:41 |
| *** rloo has joined #openstack-ironic | 12:42 | |
| lucasagomes | but changing totally unrelated stuff in the same patch I don't usually like | 12:42 |
| romcheg | maybe it was commited by mistake :D | 12:42 |
| lucasagomes | haven't reviewed so idk if it's related or not, if not there's no harm in -1 and asking to split in a new patch | 12:42 |
| lucasagomes | romcheg, possible | 12:42 |
| *** yuriyz has quit IRC | 12:44 | |
| *** yuriyz has joined #openstack-ironic | 12:46 | |
| rloo | romcheg: wrt 83572, did you read in the commit, why they changed test_ports? | 12:46 |
| *** linggao has joined #openstack-ironic | 12:46 | |
| romcheg | rloo: it's possible to move that into a separate patch anyway. It will just be a dependency for this one | 12:48 |
| rloo | romcheg: yes, that's probably true. Sorry, it did occur to me but I'm not such a stickler for separate patches so I let it go ;) | 12:51 |
| romcheg | rloo: That's just my opinion. If the rest of the team disagrees with me, I don't mind letting it go. That's how discussions work :) | 12:52 |
| rloo | romcheg: ha ha. So, to be fair about it (based on previous reviews etc) you are right. It should probably be put in a separate patch ;) | 12:52 |
| *** nosnos has quit IRC | 12:56 | |
| rloo | romcheg, yuriy, lucasagomes: https://review.openstack.org/#/c/83399/. If two cores +2, but don't approve, does a 3rd core need to review and approve or just approve w/o looking? | 13:01 |
| *** yuriyz has quit IRC | 13:01 | |
| *** yuriyz|2 has joined #openstack-ironic | 13:01 | |
| rloo | oops, yuriyz|2 ^^ | 13:02 |
| lucasagomes | rloo, I usually review and then approve | 13:02 |
| lucasagomes | I don't approve things I haven't reviewed | 13:02 |
| lucasagomes | also worth take a look at the comments | 13:02 |
| lucasagomes | sometimes people +2 but specifically say that he want to give others the oportunity to review as well | 13:02 |
| lucasagomes | he/she | 13:03 |
| rloo | i don't see that in this review. just a comment from roman. Just wondering if diff cores have diff ways of working? | 13:03 |
| rloo | ok, i'll quickly review it since I had looked at it before. | 13:04 |
| lucasagomes | rloo, I c, yeah we probably do have diff ways, but that's good as well | 13:04 |
| lucasagomes | rloo, ack thanks :D | 13:04 |
| *** overlayer has joined #openstack-ironic | 13:05 | |
| *** matsuhashi has quit IRC | 13:06 | |
| *** ilives has quit IRC | 13:07 | |
| *** ilives has joined #openstack-ironic | 13:08 | |
| NobodyCam | good morning Ironic says the man making coffee :) | 13:09 |
| *** martyntaylor has left #openstack-ironic | 13:09 | |
| rloo | g'morning NobodyCam | 13:10 |
| Mikhail_D_wk | NobodyCam: g'morning :) | 13:10 |
| romcheg | Morning NobodyCam | 13:11 |
| lucasagomes | morning NobodyCam | 13:14 |
| *** matty_dubs|gone is now known as matty_dubs | 13:14 | |
| lucasagomes | NobodyCam, now I read a bit more about the blueprint stuff that nova is doing :) | 13:14 |
| lucasagomes | NobodyCam, I liked the idea, but agree with JayF that we might not know _everything_ before actually start coding | 13:14 |
| lucasagomes | rloo, yay thanksssss :D | 13:15 |
| rloo | lucasagomes: ;) | 13:15 |
| *** Madasi has quit IRC | 13:16 | |
| *** jbjohnso has joined #openstack-ironic | 13:16 | |
| NobodyCam | morning rloo Mikhail_D_wk lucasagomes :) | 13:17 |
| *** matsuhashi has joined #openstack-ironic | 13:17 | |
| *** Madasi has joined #openstack-ironic | 13:18 | |
| NobodyCam | lucasagomes: yea, I really like the gerrit idea for BP's it was just a couple of the steps in the nova one that seemed odd to me | 13:19 |
| NobodyCam | why crate a "bad BP" | 13:19 |
| lucasagomes | NobodyCam, I think by "bad BP" they mean... create a tiny bp like 1 line comment | 13:20 |
| lucasagomes | in launchpad | 13:20 |
| lucasagomes | once the bp is approved in the specs repo, they will copy the text from specs to the bp in launchpad | 13:20 |
| lucasagomes | NobodyCam, I think it's done because | 13:21 |
| lucasagomes | theorically the bp doesn't need to be approved to u to start working on it | 13:21 |
| lucasagomes | (idk if it should be enforced tho) | 13:21 |
| lucasagomes | NobodyCam, so to keep track of the code while it's being approved, we need an entry in launchpad | 13:21 |
| lucasagomes | that's my theory for that rule at least | 13:21 |
| NobodyCam | I would rather start in the repo, once it been blessed then one of use (core) members who can actually accept the bp. cuts and pasted it to a actual BP and assigns it a priority | 13:22 |
| *** jistr has quit IRC | 13:22 | |
| lucasagomes | NobodyCam, yeah, but if someone starts working before it's approval | 13:22 |
| lucasagomes | there's no way to tag that bp | 13:22 |
| lucasagomes | if it's not in launchpad | 13:22 |
| NobodyCam | ahh | 13:23 |
| NobodyCam | I can see that | 13:23 |
| lucasagomes | yeah... it's quite sad that we are so integrated with launchpad tho | 13:24 |
| lucasagomes | that's a horrible system | 13:24 |
| NobodyCam | :) | 13:26 |
| rloo | lucasagomes, NobodyCam: I only skimmed some of that. Maybe I'm an optimist, but I interpreted "bad BP" == initial version/submission of a BP. Not necessarily bad. OTOH maybe the author really meant that. | 13:26 |
| *** jistr has joined #openstack-ironic | 13:26 | |
| lucasagomes | rloo, yeah I think ur correct | 13:26 |
| lucasagomes | rloo, bad BP is a bad way to describe it | 13:26 |
| NobodyCam | :) | 13:28 |
| NobodyCam | I can see many BP's that just say "SEE REVIEW #123" | 13:29 |
| lucasagomes | lol | 13:30 |
| *** newell has joined #openstack-ironic | 13:30 | |
| NobodyCam | lucasagomes: the lol comes then there is another patch tha link that BP to itself. | 13:32 |
| NobodyCam | s/then/when/ | 13:32 |
| lucasagomes | haha oh boy | 13:32 |
| NobodyCam | :) | 13:32 |
| lucasagomes | yeah it's kinda fucked up the way we track bps now | 13:32 |
| lucasagomes | we def need to improve it | 13:32 |
| NobodyCam | but I really like the bp review that perfect | 13:33 |
| NobodyCam | thats* | 13:33 |
| lucasagomes | +1 yeah that's much more in control | 13:34 |
| rloo | speaking of blueprints... https://blueprints.launchpad.net/ironic/+spec/nodeless-vendor-passthru | 13:36 |
| *** jgrimm has quit IRC | 13:37 | |
| rloo | lucasagomes, NobodyCam. I don't actually know what/if there is a process. Should I be reviewing code associated with a blueprint, if the blueprint doesn't have an approver or some indication that someone thinks it is OK? | 13:37 |
| lucasagomes | rloo, well there's no problem reviewing it I think | 13:38 |
| rloo | lucasagomes. There IS a problem. My TIME. | 13:38 |
| lucasagomes | I wouldn't approve without talking to the PTL | 13:38 |
| lucasagomes | rloo, heh | 13:38 |
| rloo | I'm serious. | 13:38 |
| lucasagomes | rloo, you can skip if u want | 13:38 |
| lucasagomes | rloo, sure it's fine to skip those reviews... if it's not approved, it's grand to skip it | 13:39 |
| lucasagomes | I think some people even -1 that | 13:39 |
| lucasagomes | I think I saw a case where people -1 saying the bp wasn't approved yet so they didn't want to merge | 13:39 |
| rloo | ok, will bring it up with deva then. I think we need to have clear expectations so everyone understands. | 13:40 |
| lucasagomes | rloo, +1 | 13:40 |
| lucasagomes | we need to start a wiki or something with those reviewers faq | 13:40 |
| lucasagomes | it's a lot of corner cases | 13:40 |
| rloo | well, this doesn't seem like a corner case. seems like a process or lack of process, problem? | 13:41 |
| lucasagomes | yeah indeed lack of process | 13:42 |
| lucasagomes | :( | 13:42 |
| lucasagomes | I'm about to submit a review for a blueprint that wasn't approved yet | 13:42 |
| lucasagomes | so would be good to bring it up to devananda today | 13:42 |
| rloo | ok, but you're at the end of the queue buddy :) | 13:43 |
| lucasagomes | lol | 13:43 |
| lucasagomes | yeah | 13:43 |
| NobodyCam | lucasagomes: rloo: if I see a review for a bp that is not approved I tend to want to only +/- 1 it notting in the review "would get +2 if BP was approved" | 13:47 |
| NobodyCam | that way the dev gets some feedback | 13:48 |
| rloo | NobodyCam, lucasagomes: hmm. I'll keep that in mind, but honestly, I don't plan on reviewing anything if the bp isn't approved. There are lots of other reviews that I could be looking at. | 13:50 |
| lucasagomes | NobodyCam, I c, yeah thats a good idea to use only +1 stuff | 13:50 |
| NobodyCam | :) | 13:51 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Port to oslo.messaging https://review.openstack.org/88307 | 13:59 |
| * lucasagomes done sigh | 13:59 | |
| romcheg | lucasagomes: wow | 14:02 |
| romcheg | lucasagomes: How was that? :) | 14:02 |
| lucasagomes | romcheg, heh please help review/test | 14:02 |
| lucasagomes | romcheg, tiring heh | 14:02 |
| lucasagomes | you can't change one piece and get the whole working | 14:02 |
| lucasagomes | so it's a lot of changes in diff places | 14:02 |
| romcheg | lucasagomes: Yup, I saw the same patch in nova | 14:03 |
| romcheg | It was super huge | 14:03 |
| lucasagomes | romcheg, yeah, I hope the Ironic one will be easier to review | 14:03 |
| lucasagomes | romcheg, yeah | 14:03 |
| lucasagomes | romcheg, the nova one had 78 patch-sets haha | 14:03 |
| lucasagomes | https://review.openstack.org/#/c/39929/78 | 14:03 |
| romcheg | lucasagomes: Should we do better (more)? :) | 14:04 |
| lucasagomes | romcheg, but it's better to port now than later, otherwise we are going to get stuck in the rpc stuff of the incubator and will be way more painful to port later | 14:04 |
| lucasagomes | romcheg, lol please no :P | 14:04 |
| * lucasagomes hopes for the better case, only 1 patch-set!!!! *dreams* | 14:05 | |
| lucasagomes | romcheg, damn just saw a debug thing I left | 14:05 |
| lucasagomes | ok 2 patch sets | 14:06 |
| romcheg | lucasagomes: lol | 14:06 |
| *** viktors has joined #openstack-ironic | 14:06 | |
| rloo | lucasagomes: if those config options are removed, how can ironic be backwards compatible? | 14:07 |
| romcheg | 223906 | 14:07 |
| lucasagomes | rloo, those options came from openstack/common/rpc/* stuff | 14:08 |
| romcheg | Whoops! | 14:08 |
| lucasagomes | rloo, now they exist in oslo.messaging | 14:08 |
| romcheg | You didn't see that.... | 14:08 |
| lucasagomes | but the syntax is backward compatible, the rpc.py is a wrap on oslo.messaging | 14:08 |
| lucasagomes | so it needs to understand both syntax the new and old | 14:08 |
| * romcheg entered a second factor code to the wrong place | 14:08 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Port to oslo.messaging https://review.openstack.org/88307 | 14:09 |
| lucasagomes | feck sample is not updated >.< | 14:10 |
| lucasagomes | patchset #3 | 14:10 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Port to oslo.messaging https://review.openstack.org/88307 | 14:11 |
| *** ifarkas_ has joined #openstack-ironic | 14:12 | |
| *** viktors has quit IRC | 14:13 | |
| *** ifarkas has quit IRC | 14:14 | |
| rloo | lucasagomes: thx. Just wondering if we need to document that ;) | 14:15 |
| lucasagomes | rloo, I c | 14:15 |
| lucasagomes | yeah a good bit of config options were removed now | 14:16 |
| lucasagomes | 1 new added | 14:16 |
| *** viktors has joined #openstack-ironic | 14:19 | |
| viktors | lucasagomes: you probably missed to change tools/config/oslo.config.generator.rc file in https://review.openstack.org/88307 | 14:23 |
| viktors | 14:23 | |
| lucasagomes | viktors, :( | 14:24 |
| lucasagomes | viktors, thanks I will update that :D | 14:24 |
| NobodyCam | and back.. morning romcheg :) | 14:24 |
| *** ifarkas_ has quit IRC | 14:25 | |
| viktors | lucasagomes: I just found, that you removed the large number conf options in config.sample | 14:25 |
| lucasagomes | viktors, yeah, adding it back :D thank you for pointing it out | 14:26 |
| *** ifarkas_ has joined #openstack-ironic | 14:26 | |
| viktors | lucasagomes: no problem :) | 14:26 |
| *** matsuhashi has quit IRC | 14:30 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Port to oslo.messaging https://review.openstack.org/88307 | 14:30 |
| *** jgrimm has joined #openstack-ironic | 14:33 | |
| NobodyCam | humm.. [-] Error during ClusteredComputeManager.update_available_resource: f34540e9-5cc4-420c-95f2-378fd71c7485 is not a valid node managed by this compute host. | 14:37 |
| NobodyCam | thats a strange error | 14:37 |
| agordeev | good morning/evening Ironic | 14:37 |
| NobodyCam | morning agordeev | 14:37 |
| NobodyCam | bbt..brb | 14:38 |
| lucasagomes | NobodyCam, I'm looking at it now | 14:39 |
| lucasagomes | lemme open a bug about it, I found the cause | 14:39 |
| lucasagomes | looking at how to solve | 14:39 |
| lucasagomes | NobodyCam, the cause is that u enrolled a powered on machine | 14:39 |
| Shrews | NobodyCam: quit breaking things | 14:40 |
| Shrews | :) | 14:40 |
| lucasagomes | yeah I will put a fix up soon | 14:42 |
| agordeev | jroll: JayF are you arouund? could you help me? i'm trying to test how IPA works. For now dracut has failed to boot and dropped to emergency shell. Idk how to troubleshoot it (also ironic-conductor logs are pretty silent) | 14:42 |
| lucasagomes | https://bugs.launchpad.net/ironic/+bug/1309048 | 14:42 |
| agordeev | morning NobodyCam Shrews and everybody else! | 14:42 |
| NobodyCam | lucasagomes: https://review.openstack.org/#/c/85529/15/ironic/conductor/manager.py | 14:42 |
| NobodyCam | mornign Shrews | 14:42 |
| NobodyCam | :-p | 14:42 |
| Shrews | agordeev, NobodyCam, et.al: morning | 14:42 |
| lucasagomes | NobodyCam, yeah... but I think the bug is a bit diff | 14:43 |
| NobodyCam | ok.. | 14:43 |
| lucasagomes | NobodyCam, but that would actually work around the problem | 14:43 |
| NobodyCam | that was from the last ci run of that patch | 14:44 |
| lucasagomes | NobodyCam, hmm lemme investigate a bit more, I not complete sure about what node_is_available should do | 14:44 |
| NobodyCam | lucasagomes: sure | 14:45 |
| NobodyCam | :) | 14:45 |
| agordeev | jroll: JayF also it took too long to init http://picpaste.com/pics/dracut_fail-vUpXXQTh.1397745849.png slightly less than 10 minutes. How did you test it? | 14:45 |
| *** ilives has quit IRC | 14:55 | |
| *** ilives has joined #openstack-ironic | 14:56 | |
| *** blamar has quit IRC | 14:56 | |
| *** foexle has quit IRC | 14:59 | |
| *** hemna_ has joined #openstack-ironic | 15:00 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:02 |
| *** blamar has joined #openstack-ironic | 15:04 | |
| devananda | morning, all | 15:08 |
| NobodyCam | good morning devananda :) | 15:08 |
| *** blamar has quit IRC | 15:08 | |
| NobodyCam | lucasagomes: is ^^ that releated to that error? | 15:08 |
| lucasagomes | NobodyCam, yes, that actually fixes it | 15:08 |
| lucasagomes | NobodyCam, I just dunno why we need node_is_available | 15:09 |
| lucasagomes | devananda, morning | 15:09 |
| agordeev | devananda: morning | 15:09 |
| lucasagomes | devananda, deva about bps, what you feel about? https://blueprints.launchpad.net/ironic/+spec/oslo-messaging | 15:09 |
| agordeev | lucasagomes: morning | 15:09 |
| lucasagomes | devananda, I just put a patch up doing that ^, can you approve the bp for me please? | 15:10 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:10 |
| NobodyCam | lol ... lucasagomes: stop that...hehehe | 15:11 |
| lucasagomes | NobodyCam, heh sorry, just tagged the bug in the commit message | 15:11 |
| NobodyCam | hehehe its all good.. just giving you a hard time | 15:12 |
| NobodyCam | :-p | 15:12 |
| lucasagomes | heh | 15:13 |
| lucasagomes | but I'm wondering whether we should overwirte the node_is_available it's just duplicating the tests | 15:13 |
| lucasagomes | the default implementation of node_is_available calls get_available_nodes() and check if the node is there | 15:14 |
| lucasagomes | that's grand, I think that's how it should be | 15:14 |
| lucasagomes | we shouldn't duplicate the conditions there (although it's more efficient to get() a specific node) | 15:15 |
| lucasagomes | maybe I can isolate the conditional for both methods | 15:16 |
| lucasagomes | 1 sec... NobodyCam will mark that ^ as WIP | 15:16 |
| devananda | lucasagomes: done. i need to go through all the BPs and look at the ones we should approve pre-summit | 15:17 |
| lucasagomes | devananda, ta much! | 15:17 |
| lucasagomes | devananda, I did everything in one commit, because I really dunno hw to split that in multiple ones | 15:17 |
| devananda | heh | 15:18 |
| NobodyCam | ya they both are check the exact same things | 15:18 |
| NobodyCam | *checking | 15:18 |
| lucasagomes | NobodyCam, yeah it does | 15:19 |
| devananda | lucasagomes: i haven' tlooked yet -- will that supercede my update https://review.openstack.org/#/c/87328/ ? | 15:19 |
| JayF | agordeev: I don't understand what you're saying or what that screenshot is showing me. More information? | 15:20 |
| lucasagomes | devananda, ops, it will | 15:21 |
| lucasagomes | devananda, actually I removed the whole openstack/common/rpc/ module | 15:21 |
| JayF | agordeev: as far as testing the agent, we have a couple of hardware labs we're working it, within the next 2 weeks we're going to work on a way to better test the agent locally, but we're just trying super hard to get a working prototype by Atlanta. | 15:21 |
| devananda | lucasagomes: great | 15:21 |
| NobodyCam | JayF: I cant wait to see your demo.. | 15:22 |
| JayF | I can't wait for us to have a demo | 15:22 |
| JayF | lol | 15:22 |
| *** ifarkas_ has quit IRC | 15:23 | |
| NobodyCam | :) | 15:23 |
| lucasagomes | - check-tripleo-ironic-seed-precise http://logs.openstack.org/07/88307/4/check-tripleo/check-tripleo-ironic-seed-precise/989d0b1 : SUCCESS in 36m 31s!!!!! | 15:23 |
| lucasagomes | on the olso messaging! | 15:23 |
| devananda | lucasagomes: \o/ | 15:24 |
| agordeev | JayF: hello | 15:24 |
| NobodyCam | lucasagomes: w00t | 15:26 |
| lucasagomes | :D was waiting for that check ^^ | 15:26 |
| agordeev | JayF: it looks like agent image is unable to boot properly. Then it just gives an emergency shell that's all. | 15:26 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:26 |
| lucasagomes | NobodyCam, done ^ | 15:26 |
| * NobodyCam refreshed his window | 15:26 | |
| Shrews | devananda: possibly a dumb question, but is the intent that preserve_ephemeral will only be supported with pxe? | 15:27 |
| agordeev | JayF: does it require any additional configuration at ironic's side? Maybe i'm just missing something | 15:27 |
| JayF | agordeev: that's strange, we have the agent booting on several machines. Are you building from latest master? (imagebuild/coreos/coreos_oem_inject.py is using 197.0.0) | 15:27 |
| devananda | Shrews: nope. it should be supportable by other deploy drivers, too | 15:27 |
| JayF | agordeev: the agent should just boot to a prompt from the perspective of the terminal | 15:27 |
| JayF | agordeev: the terminal is mostly useless in coreos though, as no password is set on the core user | 15:27 |
| lucasagomes | NobodyCam, argh ops, spotted an error | 15:27 |
| lucasagomes | >.< | 15:27 |
| agordeev | JayF: yup, also i'd applied your latest patch which reduces image size | 15:27 |
| devananda | hey guys -- look :) https://launchpad.net/ironic/icehouse/2014.1 | 15:28 |
| Shrews | devananda: ok. saw that pxe was the only driver with a driver_info attribute for that, so it made me question it's use | 15:28 |
| devananda | all the things we did in icehouse, in one page | 15:28 |
| devananda | Shrews: pxe is the only deply driver in trunk today, too :) | 15:28 |
| NobodyCam | \o/ | 15:28 |
| JoshNang | woo! | 15:28 |
| JayF | agordeev: if you can give me a full boot output when it fails (serial console is good for this), I'll gladly take a look, but without that information it's hard to troubleshoot | 15:28 |
| agordeev | JayF: nice, so it's impossible to figure out what happened | 15:28 |
| JayF | agordeev: no, not impossible. You can use the terminal from that rescue prompt to get info on why it wouldn't boot from dmesg and journalctl | 15:29 |
| NobodyCam | our first release... awesome work EVERYONE!!!! | 15:29 |
| agordeev | JayF: ok, i'll try to get serial console output | 15:29 |
| lucasagomes | NobodyCam, heh I forgot it was in the nova driver was running tests with testr instead of tox so I didn't see the error | 15:29 |
| JayF | agordeev: just saying if it boots to a prompt, you can't login, but that's a good thing | 15:29 |
| JayF | agordeev: aha, what's the kernel params you're passing? | 15:30 |
| JayF | I bet I just thought of what your problem is | 15:30 |
| JayF | two things: how much ram is on that box, and can you send me the ipxe.txt or pxe config file you're booting it with | 15:31 |
| agordeev | JayF: default params just what ironic has | 15:31 |
| JayF | I've never used the built in ironic dhcp/pxe stuff, we've been managing dhcp/pxe outside of ironic | 15:31 |
| JayF | I suspect it's two things: | 15:31 |
| agordeev | JayF: i'd tryed 3G and 4G RAM. It doesn't matter. What is the minimum? | 15:31 |
| JayF | 1) You need to have the correct kernel params on the command line for it to boot (I'll gist what we're using in a sec) | 15:31 |
| JayF | 2) The default run.sh for coreos is set to remount the tmpfs to be 20G instead of their default of 2G. I'd suggest you lower that number to 1/2 your ram (or comment out the line) in imagebuild/coreos/oem/run.sh | 15:32 |
| agordeev | JayF: http://paste.openstack.org/show/76140/ here is the pxe config from ironic. | 15:33 |
| JayF | https://gist.github.com/jayofdoom/c30538662b9661141f0b as I note in there, you probably don't need ipa-advertise-host among other things -- we have multiple interfaces that DHCP in our hardware, so we have to tell the agent which one to advertise on | 15:34 |
| JayF | agordeev: yeah, that append line will have to be different for the agent, and not just because we're not using DIB | 15:35 |
| *** lynxman has quit IRC | 15:36 | |
| lucasagomes | thierry is spamming my mail inbox ;( | 15:36 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:37 |
| *** lynxman has joined #openstack-ironic | 15:37 | |
| agordeev | JayF: also question. Is that information stated anywhere in docs? I think it's worth to be mentioned somewhere | 15:38 |
| JayF | agordeev: I agree. For now, I think a sample set of pxe configs in the imagebuild/coreos/ directory might be sufficient, with obviously something more significant in the future | 15:39 |
| NobodyCam | OH! -- POWER_MANAGER=nova.virt.baremetal.virtual_power_driver.VirtualPowerManager | 15:39 |
| JayF | right now you still have to apply about a half dozen patches or more to Ironic to even get the agent talking to Ironic at all, we'd like to get something working and merged before we write a ton of docs about how to use it :) | 15:40 |
| JayF | agordeev: https://github.com/openstack/ironic-python-agent/blob/master/imagebuild/coreos/oem/run.sh#L8 this is the place where you'll have to either comment out that line or change the ram amount in order to get it to boot on regular machines | 15:40 |
| *** lsmola has quit IRC | 15:41 | |
| *** eghobo has joined #openstack-ironic | 15:41 | |
| JayF | agordeev: http://coreos.com/docs/running-coreos/bare-metal/booting-with-pxe/ the section "Adding a Custom OEM" here describes what the imagebuilder is doing if you wanted more docs for now about how it works | 15:42 |
| agordeev | JayF: ironic's patches have been already applied :) | 15:42 |
| agordeev | JayF: great! | 15:42 |
| lucasagomes | Shrews, all that pxe_root_gb, pxe_swap_gb, pxe_ephemeral_gb, pxe_preverse_ephemeral... will be moved from driver_info to instance_info and should be driver agnostic | 15:42 |
| JayF | agordeev: tl;dr of ^ is that the oem/ dir gets injected into the provided coreos pxe image, and coreos runs /usr/share/oem/run.sh on boot to configure things | 15:42 |
| JayF | agordeev: in our case, that's making the tmpfs larger (optional, and something that we should probably un-upstream), injecting SSH keys, then loading up the agent container and adding the unit file to systemd | 15:43 |
| Shrews | lucasagomes: is there a current review for that? | 15:43 |
| lucasagomes | Shrews, not yet, I think one of the rackspace guys are moving that (I think there's a bug or bp about it up) | 15:43 |
| JayF | yeah, there's a BP up for it | 15:44 |
| JayF | and I think it's next up on our todo once some of the outstanding stuff gets reviewed or merged | 15:44 |
| agordeev | yuriyz|2: ^^ note about run.sh | 15:45 |
| lucasagomes | ack | 15:45 |
| lucasagomes | JayF, thanks | 15:45 |
| *** ndipanov has quit IRC | 15:49 | |
| *** dtantsur|launch is now known as dtantsur | 15:50 | |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Deprecated extension name attribute https://review.openstack.org/87557 | 15:50 |
| *** ndipanov has joined #openstack-ironic | 15:50 | |
| Shrews | biab. change of venue... | 15:51 |
| *** ndipanov has quit IRC | 15:58 | |
| rloo | devananda: when you have a moment, would you please take a look/approve this BP: https://blueprints.launchpad.net/ironic/+spec/nodeless-vendor-passthru | 15:59 |
| lucasagomes | tools/config/generate_sample.sh is returning that it's not updated from master | 16:00 |
| JayF | rloo: code already exists for that too... here --> https://review.openstack.org/#/c/81919/ | 16:00 |
| lucasagomes | it's causing pep8 gate checks to fail | 16:01 |
| JayF | rloo: that's one of the patches that's required for getting the agent driver merged, so I'm glad you're looking at the bp | 16:01 |
| rloo | JayF: yes, I know. I was about to review it, but the BP hasn't been approved. | 16:01 |
| lucasagomes | devananda, ^ should we have a patch updating the config?! | 16:01 |
| JayF | rloo: :D Thanks | 16:01 |
| devananda | lucasagomes: hm? | 16:01 |
| rloo | JayF: don't thank too soon. I won't start before the BP is approved. Is that the first review to look at? I know you guys have several. | 16:01 |
| lucasagomes | devananda, the tools/config/check_uptodate.sh is failing on master | 16:01 |
| devananda | lucasagomes: yikes. that can happen when oslo libs change | 16:02 |
| lucasagomes | devananda, :( should we make a patch fixing that? | 16:02 |
| lucasagomes | devananda, it's causing pep8 checks in gate to fail | 16:02 |
| devananda | yes | 16:03 |
| devananda | i'm testing now | 16:03 |
| lucasagomes | devananda, ack, will prepare one after u confirm | 16:03 |
| *** viktors has quit IRC | 16:03 | |
| jroll | rloo: the first patch we need merged is https://review.openstack.org/#/c/81919/ | 16:04 |
| jroll | rloo: then this is the big agent driver patch: https://review.openstack.org/#/c/84795/ | 16:05 |
| JayF | Don't we need swift urls before the big agent driver patch? | 16:05 |
| jroll | rloo: after that there's some related things | 16:05 |
| jroll | JayF: I thought that landed? | 16:05 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Update ironic.conf.sample after oslo changes https://review.openstack.org/88352 | 16:05 |
| NobodyCam | rloo: didn't you have a issue with generate_sample.sh and the keystone conf settings? | 16:05 |
| rloo | jroll: ok, so need the BP to be approved first. | 16:05 |
| jroll | JayF: oh, you're right | 16:05 |
| jroll | rloo: ah, yes | 16:06 |
| jroll | thanks | 16:06 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Update ironic.conf.sample https://review.openstack.org/88353 | 16:06 |
| rloo | NobodyCam: hmm, I don't remember. Except for the issue that generate_sample.sh doesn't generate a .sample that is quite right for some options. | 16:06 |
| jroll | rloo: devananda: this bp is related to all of the subsequent patches: https://blueprints.launchpad.net/ironic/+spec/agent-driver | 16:06 |
| devananda | FWIW nova moved away from having a conf.s.ample in tree at all, because of this problem | 16:06 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 2 https://review.openstack.org/73256 | 16:08 |
| lucasagomes | devananda, hmm... I don't know if moving away, but maybe the check shouldn't be mandatory to pass | 16:08 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 3 https://review.openstack.org/86536 | 16:08 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 1 https://review.openstack.org/73223 | 16:08 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Get rid of the newline "\" https://review.openstack.org/66793 | 16:08 |
| NobodyCam | hummm.. I like the sameple file, but if its going to cause issues :( | 16:09 |
| lucasagomes | devananda, then we could have a jenkins job that generate it for us | 16:09 |
| lucasagomes | from time to time, like once a week it generates a patch updating the sample file | 16:09 |
| devananda | i prefer the gate check | 16:09 |
| devananda | requires patch authors & reviewers to know that they're changing the config | 16:10 |
| devananda | but yea, it causes issues like this when upstream req's change, like oslo libs or keystoneclient | 16:10 |
| lucasagomes | right, then we have to deal with this out of sync problems | 16:10 |
| lucasagomes | yeah | 16:10 |
| *** rloo has quit IRC | 16:11 | |
| *** romcheg1 has joined #openstack-ironic | 16:11 | |
| devananda | russell_h: so it's a bit backwards at this point, but could you update the nodeless-vendor-passthru BP with some implementation details? | 16:11 |
| * NobodyCam git clones more coffee | 16:12 | |
| devananda | russell_h: like what the new API (would) look like, how the RPC routing (would be) is done, etc | 16:12 |
| *** rloo has joined #openstack-ironic | 16:12 | |
| *** todd_dsm has quit IRC | 16:12 | |
| devananda | russell_h: this is a problem for most of our BPs, in my opinion - -they're too light on the proposed implementation details | 16:13 |
| devananda | i get that, often, one can't know exactly how the change will play out, but eg. "the API will randomly select a suitable conductor to send the message to" | 16:13 |
| devananda | vs "the API will braodcast the message to all conductors with that driver" | 16:13 |
| *** foexle has joined #openstack-ironic | 16:17 | |
| lucasagomes | romcheg, re 83572... although I think it can be splitted in another patch, I think the changes in the manager and driver factory stuff caused test_ports to need to be updated | 16:19 |
| lucasagomes | romcheg, so i think I'm ok with that patch | 16:19 |
| romcheg1 | lucasagomes: ok, I'm going to change my vote then | 16:20 |
| lucasagomes | romcheg1, ack, but ur not wrong tho... I think yeah it could be in another patch | 16:20 |
| rloo | lucasagomes: but to be devil's advocate, test_ports can be updated first, and then the changes in mgr/driver factory will work. | 16:20 |
| lucasagomes | rloo, yup, it could | 16:21 |
| lucasagomes | rloo, but it's hard to distinct what in that patch is a refactor or needed to make it work | 16:21 |
| lucasagomes | rloo, so, I think I'm fine with that change | 16:21 |
| rloo | I'm actually fine with it, BUT I've seen previous reviews where it was suggested that things be separated... | 16:21 |
| romcheg1 | Since there are 2 +2s already I'm going to approve it | 16:22 |
| rloo | thx romcheg1 ;) | 16:22 |
| romcheg1 | my pleasure :) | 16:22 |
| lucasagomes | romcheg1, thanks! | 16:22 |
| dtantsur | lucasagomes, romcheg1, rloo: just to remind you guys: we already have 9 (!) refactoring patches around, and I've no idea how to merge them :) | 16:22 |
| rloo | dtantsur: yes, I know :-( | 16:23 |
| lucasagomes | dtantsur, :( | 16:23 |
| lucasagomes | devananda, romcheg NobodyCam rloo ... what about a review jam to get those merged ^ | 16:23 |
| lucasagomes | focus on that refactoring stuff and get what's possible in | 16:23 |
| *** Guest78217 is now known as mgagne | 16:23 | |
| *** mgagne has joined #openstack-ironic | 16:24 | |
| rloo | to be honest, I don't really think a review jam is necessary. I think if we do one at a time, and by 'do', need two cores to approve fairly quickly, it can be done. | 16:24 |
| romcheg1 | Got to rush to my classes. lucasagomes, will give you some feedback on oslo.messaging after that | 16:24 |
| dtantsur | what I want to do actually is the following: 1. make sure https://review.openstack.org/#/c/73223/ is fixed (work with Mikhail_D_wk) 2. one morning get to guys with +2 power to quickly review and approve it 3. repeat with the next one | 16:24 |
| lucasagomes | romcheg1, right, ta much for review it! | 16:24 |
| dtantsur | * get 2 guys with +2 power even | 16:25 |
| *** matty_dubs is now known as matty_dubs|lunch | 16:25 | |
| rloo | but I don't know if any of the refactoring reviews are ready. Haven't looked recently but they seemed to need jenkins approval or they depended on some other review. | 16:25 |
| devananda | I'd like to see comstud's refactoring land, as that's solving some things I feel are significant | 16:25 |
| jroll | devananda: so maybe it's just an openstack process thing. but, I don't understand the point of updating blueprints with implementation details after patches are up. the implementation details are in the code - I don't understand why someone would go back and look at completed blueprints (and thus why do we need to update those?). | 16:25 |
| devananda | jroll: yea. folks do go back and look at BPs as artefacts of process by which they judge what has changed in the project at release time | 16:26 |
| rloo | jroll: from my perspective, I don't want to find out what something is supposed to do when I'm reviewing it. I'd like to see some sort of design first. But I may be a minority in this. | 16:26 |
| devananda | jroll: taht said, to a degree, it's process overhead taht I don't like | 16:26 |
| jroll | devananda: but do they look at implementation details to judge what has changed? | 16:26 |
| devananda | jroll: and i'd rather we got those details sorted up front (during BP review, before a single line of code is written | 16:26 |
| devananda | jroll: so i'd ilke us to start movnig towards the specs repo style of work that nova, neutron ,and tripleo are adopting | 16:27 |
| devananda | jroll: but not going to impose taht on existing work ... | 16:27 |
| comstud | +1 on specs repo | 16:27 |
| devananda | jroll: and yes, there are point-haired-folks who DO look at the BPs and DONT look at the code | 16:27 |
| devananda | jroll: and they make judgements based on that | 16:27 |
| jroll | devananda: I know you're busy, but this blueprint has sat for two months with no questions or comments until now when people are ready to land it. | 16:27 |
| jroll | devananda: sigh | 16:27 |
| devananda | jroll: i know... sorry :( | 16:28 |
| jroll | I understand | 16:28 |
| jroll | I'm just getting frustrated at constantly rebasing things while code is being changed out from under us, sorry if I'm coming off unpleasant :/ | 16:28 |
| devananda | jroll: what i'm asking for now is that you add some description of the change to the BP, so that there is an artefact describing it | 16:29 |
| jroll | right | 16:29 |
| devananda | jroll: no worries. your frustration is justified, and fairly directed at me. I can take it :) | 16:29 |
| jroll | heh, thanks :) | 16:29 |
| jroll | (for putting up with me) | 16:29 |
| devananda | jroll: I'd say that the core team wasn't large enough 2 months ago to handle the influx of patches that you guys have added -- but I also believe I said that we propbably wouldn't get to looking at the agent code until Juno opened | 16:30 |
| devananda | jroll: and the core team just expanded and we're trying to catch up now :) | 16:31 |
| jroll | devananda: I know, that's fair | 16:31 |
| * jroll will poke russell_h to update the blueprint | 16:33 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Implement caching for master images https://review.openstack.org/85387 | 16:34 |
| *** dshulyak has quit IRC | 16:35 | |
| comstud | jroll: What's the conductor method being called in this thread starvation issue? change_node_power_state ? | 16:36 |
| *** dshulyak has joined #openstack-ironic | 16:36 | |
| jroll | comstud: yes | 16:37 |
| comstud | ty | 16:37 |
| comstud | what does task.driver.power.validate do? ipmi call? | 16:38 |
| jroll | yes, checks power status | 16:39 |
| jroll | I believe | 16:39 |
| comstud | ty | 16:39 |
| * jroll checks | 16:39 | |
| comstud | i assume task.driver.power == ipmitool driver? | 16:39 |
| comstud | yeah, IPMIPower, I'm guessing | 16:40 |
| jroll | yeah, it does an IPMI call to make sure the bmc is there | 16:40 |
| comstud | ok | 16:40 |
| jroll | yes, ipmitool | 16:40 |
| lucasagomes | comstud, check if the credentials you have is valid | 16:40 |
| comstud | thanks | 16:40 |
| *** rloo has quit IRC | 16:47 | |
| JayF | agordeev: https://bugs.launchpad.net/ironic/+bug/1309110 you /do/ have to set ipa-advertise-host. I filed a bug to make that default much better. | 16:48 |
| *** rloo has joined #openstack-ironic | 16:48 | |
| JayF | agordeev: would be a very good low hanging fruit for someone wanting to get a start in the agent. | 16:48 |
| *** jistr has quit IRC | 16:49 | |
| *** zdin0bot has joined #openstack-ironic | 16:49 | |
| lucasagomes | done for today, came to the office so gotta take the train back home | 16:55 |
| NobodyCam | have a good ride back home | 16:55 |
| lucasagomes | have a g'night devananda NobodyCam rloo comstud JayF jroll , everyone... | 16:55 |
| comstud | night! | 16:55 |
| jroll | night lucas | 16:55 |
| rloo | ciao lucasagomes! | 16:55 |
| dtantsur | lucasagomes, g'night | 16:56 |
| JayF | night | 16:56 |
| comstud | jroll: what state is passed to change_node_power_state? | 16:57 |
| comstud | Is this REBOOT, power on, off, or.. ? | 16:57 |
| jroll | I was mostly sending off | 16:57 |
| comstud | ok, ty | 16:58 |
| jroll | but have seen it with on | 16:58 |
| *** lucasagomes has quit IRC | 17:00 | |
| *** eghobo has quit IRC | 17:02 | |
| *** derekh has quit IRC | 17:04 | |
| *** max_lobur has quit IRC | 17:04 | |
| *** eghobo has joined #openstack-ironic | 17:07 | |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Better documentation and defaults for coreos image https://review.openstack.org/88364 | 17:07 |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Use docker import/export to make image smaller https://review.openstack.org/87819 | 17:07 |
| jroll | agordeev: I know it's kinda of late for this, but how would y'all feel about writing up a blueprint for partioning in the agent? I'd kind of like to see the end goal - what is planned to be supported, not supported, etc. | 17:07 |
| *** dwalleck has joined #openstack-ironic | 17:07 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 17:07 | |
| vkozhukalov | Guys, what are you thinking about assigning more people to be cores in IPA in European timezone? I mean some core reviewers romcheg and yzveryanskyy who are in european tz. I believe it can help to speed up merging patches. | 17:08 |
| vkozhukalov | afaik, lucasgomes is also in Europe | 17:09 |
| jroll | vkozhukalov: my first question is, do you see problems with the speed of reviews? | 17:09 |
| jroll | or do you just mean that a patchset cycle takes a full day because we don't review until you're done for the day? | 17:10 |
| vkozhukalov | actually, I see the problem with merging particular request https://review.openstack.org/#/c/86163/ | 17:10 |
| jroll | I'm open to adding some ironic cores to IPA core team if needed, we even talked about merging the two teams at one point | 17:11 |
| jroll | vkozhukalov: right, so it sucks that it takes a full day to do a patchset / review cycle due to timezones | 17:11 |
| vkozhukalov | jroll: exactly | 17:12 |
| jroll | but we have reviewed each patchset on that review quickly | 17:12 |
| jroll | ok | 17:12 |
| jroll | JayF / JoshNang / russell_h wdyt? | 17:12 |
| JayF | I think some of what will help with this is the design summit at Atlanta and better up front understanding about what's being done | 17:12 |
| *** harlowja_away is now known as harlowja | 17:12 | |
| JayF | on that merge request in particular, I think it's awkward to review because it's a middle step without a clear idea of how it's going to integrate in the end | 17:13 |
| vkozhukalov | jroll: yes, i don't think that review is slow, you guys are really rock, I just think that we can improve the situation a bit | 17:13 |
| JayF | i.e. all the existing stuff in the agent assumes that an image with partitions is being given in, how does the agent determine what disks to partition, etc | 17:13 |
| JayF | so for me personally, even though I've already +2'd it, it probably wasn't a good idea to in retrospect because I don't know what the next steps are | 17:14 |
| JayF | I'm hoping that once we have something working, and the base layer merged, we'll all be able to talk about where we're going, and things like that | 17:15 |
| JayF | the main piece that I think hasn't been well communicated is the scope of the agent, and there might be some conflict about what things belong in the agent vs not (like the LVM/MD conversation we had with devananda the other day) | 17:15 |
| JayF | and until the scope is well understood by all, I'd be nervous about expanding the core group of reviewers | 17:16 |
| vkozhukalov | there is bp https://blueprints.launchpad.net/ironic/+spec/ironic-python-agent-partition and everyone is welcome to criticize it | 17:16 |
| jroll | heh, I had no idea this existed | 17:16 |
| vkozhukalov | next step is making paritiotion extension | 17:17 |
| jroll | vkozhukalov: I'd like to see more info on that - what is supported, what is not supported, what the goal is, etc | 17:17 |
| vkozhukalov | jroll: JayF: it is my fault, that you did not know about it | 17:17 |
| jroll | that's ok | 17:17 |
| vkozhukalov | https://etherpad.openstack.org/p/ironic-disk-partitioning some additional info | 17:18 |
| comstud | devananda, jroll: Ok. So, I think I know what is happening with regards to the thread starvation | 17:18 |
| *** rloo has quit IRC | 17:19 | |
| JayF | vkozhukalov: can you reflect the contents of that etherpad in the blueprint? or at a minimum link it in? | 17:19 |
| vkozhukalov | it is clear for now about lvm, we are not going to support it in agent, but md is still in the scope | 17:19 |
| *** rloo has joined #openstack-ironic | 17:19 | |
| JayF | AFIACT, it's not in scope for the same reason LVM isn't. | 17:19 |
| comstud | devananda, jroll: What is happening is that we get into a state where all RPC pool threads are blocked waiting on the Conductor worker threads. And in this state, the scheduling list is such that none of the RPC pool threads run until *all* of the Conductor workers complete. | 17:19 |
| vkozhukalov | there is a link in bp to this etherpad | 17:19 |
| comstud | devananda, jroll: So nothing is sending RPC responses until all of the worker threads are done... and at a default of 64, this can be a long time. | 17:21 |
| devananda | comstud: hmm... taht doesn't seem logical | 17:21 |
| vkozhukalov | JayF: I don't remember anyone was against md. | 17:21 |
| comstud | devananda, jroll: It's how eventlet does scheduling.. which is arguably bad. | 17:21 |
| jroll | side note, all involved with the agent, I've updated the main agent blueprint (and would like to see blueprints for other agent features in general): https://blueprints.launchpad.net/ironic/+spec/utility-ramdisk | 17:21 |
| devananda | comstud: ah. lovely | 17:21 |
| devananda | comstud: is taht only a problem if all worker threads get consumed? | 17:21 |
| JayF | Hmm I thought the argument was that if you couldn't use it with a Windows image, you shouldn't support it | 17:21 |
| devananda | fwiw guys, i just emailed the list re: an ironic specs repo | 17:21 |
| comstud | devananda, jroll: Say your worker thread pool is full. So you have RPC threads blocking on them.. | 17:21 |
| JayF | because of the strangeness of, should we support /windows/ software raid as well? | 17:22 |
| devananda | JayF: fwiw, I do not beleive i said that | 17:22 |
| comstud | devananda, jroll: When a worker finishes... it doesn't immediately schedule the RPC thread | 17:22 |
| JayF | There's just a whole path of nastiness that supporting linux software raid takes us down | 17:22 |
| comstud | devananda, jroll: It just adds to end of the list | 17:22 |
| JayF | devananda: my misunderstanding then | 17:22 |
| comstud | devananda, jroll: So that means all of the workers have to be scheduled and run first before you get back to the RPC pool thread | 17:22 |
| *** matty_dubs|lunch is now known as matty_dubs | 17:22 | |
| jroll | comstud: sure... the thread I see getting starved is a heartbeat looping call thing. where does that fit in? | 17:22 |
| comstud | ah right, I did't factor that in here | 17:22 |
| jroll | :) | 17:22 |
| *** rloo has quit IRC | 17:22 | |
| comstud | I was focusing on API locking up | 17:22 |
| vkozhukalov | we agreed that we can store some attributes in image metadata about supporting or not such things like md | 17:23 |
| jroll | comstud: I mean, that's a problem too :P | 17:23 |
| comstud | of which there's a clear explanation | 17:23 |
| *** rloo has joined #openstack-ironic | 17:23 | |
| comstud | but | 17:23 |
| JayF | vkozhukalov: that etherpad is not linked in the blueprint here: https://blueprints.launchpad.net/ironic/+spec/ironic-python-agent-partition | 17:23 |
| comstud | the periodic task would have the same problem here... | 17:23 |
| JayF | Oh! I see it now! | 17:23 |
| JayF | vkozhukalov: ^ sorry, missed the "read the full specification" | 17:23 |
| comstud | It can be behind all of the worker threads and need to wait for them all to complete | 17:23 |
| comstud | but it certainly should run | 17:23 |
| *** rloo has quit IRC | 17:24 | |
| jroll | right, it does run... about 20x slower than it should :) | 17:24 |
| comstud | yeah | 17:24 |
| *** rloo has joined #openstack-ironic | 17:24 | |
| comstud | if 64 ipmi + DB calls takes 2 minutes... | 17:24 |
| comstud | the periodic task is going to be... behind | 17:25 |
| jroll | yeah | 17:25 |
| comstud | (64 being the default pool size) | 17:25 |
| comstud | devananda, jroll: So the solution here is rpc pool size of 4 or something small. :) | 17:25 |
| comstud | for now | 17:25 |
| jroll | makes sense | 17:25 |
| jroll | I'll test that this afternoon | 17:25 |
| jroll | poke me if you don't hear from me about it :P | 17:26 |
| comstud | devananda: I was telling jroll earlier.. we found this happen with nova-scheduler and DB calls also | 17:26 |
| devananda | comstud: interesting. is this how you solved it? | 17:26 |
| comstud | devananda: 100 concurrent builds... the 100 DB calls happen to run before we do any processing on the return from the first one.. (we have an explicit yield after every DB call.. and it happened to cause this problem) | 17:26 |
| comstud | yes | 17:26 |
| comstud | we lowered down to 4 | 17:26 |
| devananda | why 4? | 17:26 |
| devananda | why not 8? | 17:27 |
| russell_h | comstud: is this just a scheduler bug? | 17:27 |
| comstud | well | 17:27 |
| comstud | I'm not sure if I would call it a bug | 17:27 |
| comstud | but it's an eventlet implementation detail | 17:27 |
| devananda | hm | 17:28 |
| comstud | devananda: well | 17:28 |
| comstud | devananda: it's a guess. but.. | 17:28 |
| comstud | say IPMI + DB call takes 1s | 17:28 |
| comstud | 8 threads means something could be waiting 8 seconds | 17:28 |
| comstud | 4 threads means something could be waiting 4 seconds | 17:28 |
| devananda | comstud: ipmi can reasonably take *much* longer | 17:29 |
| comstud | assuming those worker threads are all scheduled in a row | 17:29 |
| comstud | which they are | 17:29 |
| devananda | like 30 seconds | 17:29 |
| comstud | ok, then we may need to do something to force some better scheduling. | 17:29 |
| devananda | which is why we spin off a separate greenthread to handle that, and the RPC response should be sent back from conductor separately | 17:29 |
| comstud | right | 17:30 |
| comstud | well | 17:30 |
| devananda | but when we designed that, i dont think any of us were aware of this eventlet scheduling detail | 17:30 |
| comstud | I think ditching the pool would maybe help here.. I was going to model that out. | 17:30 |
| comstud | and see what the scheduling would look like | 17:30 |
| comstud | (ditcht he worker pool) | 17:30 |
| comstud | And just call eventlet.spawn() | 17:30 |
| *** foexle has quit IRC | 17:30 | |
| devananda | hm | 17:30 |
| comstud | because that would keep the RPC pool from blocking | 17:30 |
| comstud | however... | 17:30 |
| comstud | i dunno maybe worker pool just needs to be 2 * rpc pool size | 17:31 |
| comstud | or something | 17:31 |
| russell_h | so that would exacerbate the issue if you did wind up with a ton of threads blocked on DB calls right? | 17:31 |
| comstud | yeah, you don't really want to stack up infinite numbers of threads, probably | 17:31 |
| devananda | comstud: why would calling eventlet.spawn() vs GreenPool.spawn() be different here? | 17:31 |
| comstud | and you'll run out of resources doing a million fork()/exec for ipmitool | 17:31 |
| comstud | devananda: The RPC pool thread is blocking on the worker pool... an available thread | 17:32 |
| russell_h | is there a reasonable facility for putting db calls into OS threads? | 17:32 |
| comstud | devananda: so, going back on what I said.. I think maybe things are better if the worker pool is a multiple of the rpc pool | 17:32 |
| devananda | comstud: wait. it's blocking trying to get a free worker? no... it raises ane xception | 17:32 |
| comstud | lower rpc pool down... and keep worker pool higher | 17:32 |
| comstud | devananda: huh | 17:32 |
| devananda | comstud: there's an explicit limit to the worker pool size to avoid infinite resource consumption there | 17:32 |
| devananda | comstud: see manager._spawn_worker | 17:33 |
| comstud | 66 if self._worker_pool.free(): | 17:33 |
| comstud | heh | 17:33 |
| devananda | right | 17:33 |
| comstud | so i missed this | 17:33 |
| vkozhukalov | devananda: can we place this bp https://blueprints.launchpad.net/ironic/+spec/ironic-python-agent-partition as a dependency into bp tree here https://blueprints.launchpad.net/ironic/+spec/utility-ramdisk? | 17:33 |
| comstud | Well, what the hell | 17:33 |
| devananda | comstud: in my tests, it's easy to hit that exception | 17:33 |
| devananda | but the RPC messages are still not getting returned timely | 17:33 |
| comstud | I think there's another detail here I'm missing... with the RPC response | 17:34 |
| comstud | i'm assuming it's not causing a switch | 17:34 |
| comstud | but it could be | 17:34 |
| *** tatyana has joined #openstack-ironic | 17:34 | |
| comstud | either way.. The problem has to be something with the RPC pool threads... | 17:34 |
| comstud | not running until after all conductor workers finish | 17:35 |
| *** zdin0bot has quit IRC | 17:35 | |
| comstud | Somehow we're getting into a state where the scheduling list in eventlet is in that order | 17:35 |
| comstud | worker threads all in a row -> rpc pool threads all in a row | 17:35 |
| comstud | I think that's the only thing that makes sense here | 17:35 |
| *** overlayer has quit IRC | 17:36 | |
| comstud | but | 17:36 |
| devananda | vkozhukalov: done. also updated status of that bp | 17:36 |
| comstud | The fact that you raise on no workers nixes a big part of my hypothesis | 17:36 |
| comstud | :) | 17:36 |
| devananda | heh | 17:37 |
| comstud | I would try a lower rpc pool size, however... drop it down to like 4 | 17:38 |
| comstud | And see if you get API responses more quickly | 17:38 |
| comstud | and maybe raise # of workers to be a mulitple of rpc pool size | 17:39 |
| comstud | (give it its own conf setting) | 17:39 |
| *** rloo has quit IRC | 17:39 | |
| *** rloo has joined #openstack-ironic | 17:40 | |
| *** dwalleck has quit IRC | 17:40 | |
| comstud | anyway, i gotta run an errand real quick | 17:40 |
| comstud | bbiab | 17:40 |
| devananda | cheers, thanks for the investigation | 17:41 |
| devananda | i'll test those out on my other machine shortly | 17:41 |
| russell_h | devananda: filled this out some: https://blueprints.launchpad.net/ironic/+spec/nodeless-vendor-passthru | 17:41 |
| devananda | russell_h: much better, thanks! | 17:43 |
| devananda | russell_h: oh. ya'll can also mark the implementation as 'started' when you post code for a BP -- i just did it for that one | 17:44 |
| russell_h | thanks | 17:45 |
| devananda | russell_h: afaik, that field is independent of direction/definitino/etc | 17:45 |
| devananda | and can bet set by the bp author | 17:45 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Implement caching for master images https://review.openstack.org/85387 | 17:47 |
| *** zdin0bot has joined #openstack-ironic | 17:49 | |
| *** zdin0bot has quit IRC | 17:49 | |
| *** zdin0bot has joined #openstack-ironic | 17:49 | |
| JayF | FYI: I'm going to be out and mostly-unreachable from this afternoon through Tuesday morning. | 17:49 |
| dtantsur | Guys, could someone comment on the direction I am going to with https://review.openstack.org/#/c/85387/ ? (It is by no means finished or close to be finished) | 17:52 |
| dtantsur | Thanks in advance :) | 17:52 |
| devananda | huh, it looks like http://docs.openstack.org/developer/ironic/ is suddenly out of date? | 17:53 |
| devananda | or i've got a bad cache somewhere in the way | 17:54 |
| NobodyCam | humm restoreing from backup maybe? | 17:55 |
| openstackgerrit | A change was merged to openstack/ironic: Update ironic.conf.sample https://review.openstack.org/88353 | 17:55 |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Use docker import/export to make image smaller https://review.openstack.org/87819 | 17:56 |
| devananda | JoshNang: can you join #openstack-swift? | 18:01 |
| devananda | JoshNang: we're talking about the tempurl patch | 18:01 |
| JoshNang | devananda: certainly | 18:02 |
| *** ilives has quit IRC | 18:03 | |
| *** ilives has joined #openstack-ironic | 18:04 | |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk utils https://review.openstack.org/86163 | 18:09 |
| vkozhukalov | JayF: changed commit message https://review.openstack.org/#/c/86163/ and added comments into the whiteboard https://blueprints.launchpad.net/ironic/+spec/ironic-python-agent-partition | 18:10 |
| JayF | I'll make sure to have a look before I leave today -- but I'm out until Tuesday. I'll trust jroll and other cores to tkae my concerns into account if they want to +2 while I'm gone :) | 18:11 |
| *** zdin0bot has quit IRC | 18:15 | |
| vkozhukalov | JayF: ok, thanks | 18:15 |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Uniquely identify builds to prevent race https://review.openstack.org/88383 | 18:16 |
| *** rloo has quit IRC | 18:17 | |
| *** rloo has joined #openstack-ironic | 18:17 | |
| *** rloo has quit IRC | 18:26 | |
| *** rloo has joined #openstack-ironic | 18:27 | |
| *** rloo has quit IRC | 18:29 | |
| *** rloo has joined #openstack-ironic | 18:29 | |
| NobodyCam | brb | 18:29 |
| *** rloo has quit IRC | 18:31 | |
| *** rloo has joined #openstack-ironic | 18:32 | |
| *** tatyana has quit IRC | 18:33 | |
| *** vkozhukalov has quit IRC | 18:34 | |
| * devananda gets brunch | 18:48 | |
| openstackgerrit | Tushar Kalra proposed a change to openstack/ironic: cleanup docstring for drivers.utils.get_node_mac_addresses https://review.openstack.org/88395 | 18:49 |
| NobodyCam | enjoy devananda :) | 18:49 |
| Shrews | i was so deep in concentration on my work that my laptop battery just died on me. doh! | 18:53 |
| NobodyCam | doh :-p | 18:53 |
| * Shrews blames NobodyCam | 18:53 | |
| *** dwalleck has joined #openstack-ironic | 18:54 | |
| NobodyCam | heheh :) | 18:54 |
| *** dwalleck_ has joined #openstack-ironic | 18:57 | |
| *** athomas has quit IRC | 18:57 | |
| *** dwalleck has quit IRC | 18:59 | |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic: Fix bypassed reference to node state values https://review.openstack.org/88403 | 19:05 |
| comstud | back | 19:08 |
| NobodyCam | wb | 19:08 |
| *** hemna_ has quit IRC | 19:14 | |
| rloo | devananda: did you do anything to get this updated http://docs.openstack.org/developer/ironic/ ? | 19:17 |
| * rloo sees a period missing. Shouldn't have +2'd it. | 19:17 | |
| *** ilives has quit IRC | 19:22 | |
| *** rloo has quit IRC | 19:24 | |
| *** rloo has joined #openstack-ironic | 19:25 | |
| *** zdin0bot has joined #openstack-ironic | 19:25 | |
| comstud | could use a 2nd +2 on this: https://review.openstack.org/#/c/87076/ | 19:26 |
| *** hemna_ has joined #openstack-ironic | 19:27 | |
| * NobodyCam looks | 19:27 | |
| comstud | ty | 19:28 |
| *** zdin0bot has quit IRC | 19:30 | |
| devananda | rloo: i don tknow why the docs page seems to have reverted | 19:30 |
| *** rloo has quit IRC | 19:30 | |
| devananda | rloo: welp. now it's back. | 19:31 |
| *** rloo has joined #openstack-ironic | 19:31 | |
| *** rloo has quit IRC | 19:32 | |
| *** killer_prince has quit IRC | 19:32 | |
| *** rloo has joined #openstack-ironic | 19:32 | |
| *** rloo has quit IRC | 19:33 | |
| *** rloo has joined #openstack-ironic | 19:34 | |
| *** lynxman has quit IRC | 19:34 | |
| *** lynxman has joined #openstack-ironic | 19:34 | |
| NobodyCam | any reason to add a node in maintenance mode to these tests? | 19:35 |
| comstud | no, because the DB call filters it | 19:35 |
| comstud | in this patch | 19:35 |
| comstud | but | 19:35 |
| comstud | you'll see in my follow up patch.. I change this | 19:35 |
| comstud | and add a test | 19:35 |
| comstud | I mean.. the code I'm testing here has no checks on 'maintenance' mode at all in the code path. | 19:37 |
| NobodyCam | ahh so there is more to come :) (oh and I love the lambda trick) | 19:37 |
| comstud | (other than we pass 'maintenance': False to get_nodeinfo_list) | 19:37 |
| comstud | (which is tested) | 19:37 |
| *** rloo has quit IRC | 19:38 | |
| *** rloo has joined #openstack-ironic | 19:38 | |
| comstud | yeah, in my follow up patch I change code so that we're actually checking maintenance | 19:38 |
| comstud | to try to avoid races... so I add a test there | 19:38 |
| comstud | (that patch is already +2d -- but this one has to land first :) | 19:39 |
| comstud | (er +A'd) | 19:39 |
| *** killer_prince has joined #openstack-ironic | 19:40 | |
| adam_g | uh oh | 19:40 |
| adam_g | https://bugs.launchpad.net/ironic/+bug/1309156 | 19:40 |
| rloo | adam_g: uh oh. Easy to fix though. | 19:41 |
| jroll | that's.... fun | 19:41 |
| comstud | that's kind of nasty | 19:41 |
| adam_g | rloo, yeah, and backport | 19:41 |
| rloo | adam_g: not sure it needs to be backported. The change was made recently. | 19:42 |
| adam_g | oh? cool | 19:42 |
| rloo | adam_g: like maybe today. there were two places. let me check. | 19:42 |
| adam_g | is there even going to be a long-lived stable/icehouse branch of ironic? | 19:42 |
| adam_g | 828060ff79f3f96f143331e3054f66685e5f562d | 19:43 |
| rloo | adam_g: i was wrong. one of the changes hasn't gone through yet: https://review.openstack.org/#/c/84687/ | 19:43 |
| adam_g | rloo, oh, nice. will test those now | 19:44 |
| rloo | adam_g: so to fix it, that review needs to be modified to allow for power on and off. | 19:44 |
| NobodyCam | comstud: LGTM | 19:44 |
| NobodyCam | oh seems like its afternoon walkies time.. brb | 19:45 |
| rloo | adam_g: but I know there was a discussion on IRC about whether it should be power on and off, or just power off. but i didn't pay attention to it. Guess it needs to be on and off though. | 19:45 |
| comstud | NobodyCam: thnx! | 19:45 |
| rloo | adam_g: just looking at https://bugs.launchpad.net/ironic/+bug/1309048. Seems like there could be a problem if node is power on... | 19:47 |
| devananda | rloo: hi! i just realized https://review.openstack.org/#/c/73005/ got dropped | 19:48 |
| devananda | rloo: i'd love to see that re-opened -- i think lucas and I were both almost ready to land it a week ago | 19:48 |
| rloo | devananda: yeah. I didn't think it was as important as some other stuff. | 19:48 |
| rloo | devananda: ok, I'll fix it next week then. (am off in a few hours, til next Tues.) | 19:49 |
| adam_g | rloo, yeah, but that may not be an issue. eg, an admin has powered the node on to do some maintenance on it out of band from nova? | 19:49 |
| devananda | rloo: you're probably right, but it's good clean up, both in the driver code and in the API | 19:49 |
| rloo | adam_g: I just skimmed that bug, don't know the details and haven't thought it through. wanted to point it out to you just in case... | 19:49 |
| devananda | adam_g: not sure I grok that bug yet | 19:52 |
| devananda | gah | 19:53 |
| * devananda re-reads the definition of nova.virt.driver:node_is_available | 19:53 | |
| devananda | : Return whether this compute service manages a particular node. | 19:53 |
| adam_g | rloo, powering the node on seems to work fine (with that patch). removed from hypervisors resources until it is powered off again, and n-cpu restarts just fine | 19:53 |
| NobodyCam | devananda: are you also looking at https://review.openstack.org/#/c/84687/ | 19:54 |
| *** vkozhukalov has joined #openstack-ironic | 19:55 | |
| adam_g | devananda, yeah, just conflicting reports from the driver about what exists and whats available | 19:55 |
| adam_g | s/available/manageable | 19:55 |
| adam_g | 84687 seems to fix it here | 19:55 |
| rloo | adam_g: oh, so both had to be in sync wrt power off only? | 19:57 |
| rloo | adam_g: both functions. | 19:57 |
| devananda | I should have looked at / remembered how nova.compute.manager handles resources | 19:58 |
| devananda | re-reading the code now, hiding a node that is functional but provisioned doesn't seem the way that nova expects it to work | 19:59 |
| devananda | it should be presented as totally-utilized, not hidden | 19:59 |
| adam_g | rloo, right, well currently get_available_nodes() lists any node /w a power state (OFF or ON) | 19:59 |
| *** killer_prince has quit IRC | 19:59 | |
| devananda | https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L5441 | 20:00 |
| adam_g | devananda, yeah, with that patch i'm seeing nova deleting orphan compute nodes for those that are in use | 20:00 |
| devananda | right | 20:00 |
| devananda | it shouldn't be doing that | 20:00 |
| devananda | they're not orphans - -they're in use | 20:00 |
| devananda | that patch is fixing it the wrong way, IMO | 20:00 |
| adam_g | maybe the kind of availability check we're doing now should be moved into the calculation of available resources | 20:01 |
| devananda | yes | 20:02 |
| devananda | though i believe there's an open bug that the resourcemanager isn't working right now | 20:02 |
| *** zdiN0bot has joined #openstack-ironic | 20:03 | |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Better documentation and defaults for coreos image https://review.openstack.org/88364 | 20:03 |
| openstackgerrit | Thomas Bechtold proposed a change to openstack/ironic: Improve error handling for conf sample generation https://review.openstack.org/88420 | 20:04 |
| comstud | devananda: ya, that's what I was going to say.. show them to compute manager as fully utilized | 20:06 |
| comstud | maintenance ones that are not provisioned... maybe you can hide those.. although | 20:07 |
| comstud | hm, ya. | 20:07 |
| *** killer_prince has joined #openstack-ironic | 20:08 | |
| devananda | oh hum | 20:09 |
| devananda | what if a provisioned instance is put into maintenance mode in ironic... | 20:09 |
| devananda | nova is suddenly goign to NOT see that resource | 20:09 |
| devananda | but still see its instance | 20:09 |
| devananda | :( | 20:09 |
| comstud | right | 20:10 |
| comstud | which is why i think you still present it.. but fully utilized (as it would be with an instance) | 20:10 |
| comstud | else you get the bomb on nova-compute start, it appears | 20:11 |
| comstud | nova's ComputeNode table maybe should have a field for maintenance or disabled | 20:11 |
| comstud | instead of just the service record | 20:11 |
| comstud | actually, maybe it does | 20:12 |
| devananda | adam_g: should we revert 84214 while waiting for a fix? | 20:14 |
| *** jistr has joined #openstack-ironic | 20:15 | |
| adam_g | devananda, maybe? im hacking ontop of lucas' patch atm | 20:15 |
| devananda | adam_g: try http://paste.openstack.org/show/76169/ | 20:17 |
| adam_g | devananda, yeah, i have that already. that will throw resource tracking off though, working on that | 20:17 |
| devananda | k | 20:18 |
| adam_g | devananda, this is what i had in mind, http://paste.ubuntu.com/7270560/ | 20:24 |
| adam_g | not working just yet, need to run out. back in 20 | 20:24 |
| *** epim has joined #openstack-ironic | 20:25 | |
| devananda | adam_g: this n_list = icli.node.list(maintenance=False, associated=False) doesn't seem to fit | 20:25 |
| NobodyCam | oh | 20:25 |
| devananda | nova doc string says | 20:26 |
| devananda | Returns nodenames of all nodes managed by the compute service. | 20:26 |
| devananda | and update_available_resources starts with the list returned by get_available_nodes | 20:26 |
| devananda | if a node isn't in taht list, it gets deleted from the resourcetracker | 20:26 |
| adam_g | devananda, yeah--that was just a quick mod to 84687. will continue when i get back, f you haven't beat me to it :) | 20:27 |
| devananda | i actually need to step away for a while as well, probably about an hour | 20:29 |
| devananda | goign to test comstud's theory first, tho | 20:31 |
| *** romcheg1 has quit IRC | 20:32 | |
| *** jdob has quit IRC | 20:34 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 20:37 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests DO NOT MERGE https://review.openstack.org/85529 | 20:41 |
| devananda | comstud: setting rpc_thread_pool=4 and conductor.worker_pool=64, even with ~100 nodes and spamming multiple requests for each node, with a 10-second utils.execute(sleep), it's fine | 20:42 |
| devananda | jroll: ^ | 20:42 |
| devananda | taht seems to solve your conductor stops heartbeating problem | 20:43 |
| jroll | niiiice | 20:43 |
| devananda | and my API requests are not blocking either | 20:43 |
| NobodyCam | bbiafm | 20:43 |
| devananda | jroll: so to be clear, this works for utils.execute() calling ipmitool. it does not address slow DB queries | 20:44 |
| devananda | but at least locally, with 100 nodes, i'm not seeing any | 20:44 |
| jroll | devananda: right | 20:44 |
| jroll | thanks for testing that :) | 20:44 |
| devananda | jroll: i'm also seeing >30 nodes with power operations in progress at the peak of my test | 20:46 |
| devananda | there's some throttling happening... that number is quite likely an artefact of my test env | 20:46 |
| openstackgerrit | Tushar Kalra proposed a change to openstack/ironic: Remove hardcoded node id value https://review.openstack.org/88433 | 20:47 |
| *** eguz has joined #openstack-ironic | 20:49 | |
| *** rloo has quit IRC | 20:49 | |
| *** rloo has joined #openstack-ironic | 20:49 | |
| *** eghobo has quit IRC | 20:53 | |
| devananda | jroll: commented on the bug https://bugs.launchpad.net/ironic/+bug/1308680 -- are you guys going to post a fix? | 20:53 |
| jroll | devananda: yeah, would be happy to | 20:55 |
| jroll | just assigned it to myself | 20:56 |
| jroll | devananda: would you be backporting this? | 20:56 |
| jroll | probably not with a new conf option, right? | 20:56 |
| linggao | Hi ironic, why is ironic.conf.sample not up to date on the master branch? | 21:06 |
| linggao | I just git pull origin master, and tox shows that ironic.conf.sample is not up to date. | 21:07 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests DO NOT MERGE https://review.openstack.org/85529 | 21:08 |
| jroll | linggao: that checks against your local copy of ironic.conf.sample | 21:09 |
| jroll | I believe | 21:10 |
| NobodyCam | doh guess have to close parentheses when you open them :-p | 21:10 |
| jroll | oh no, you're right, linggao, my mistake | 21:10 |
| devananda | linggao: i fixed that earlier today, i believe https://review.openstack.org/#/c/88353/ | 21:10 |
| jroll | that's not good :| | 21:10 |
| devananda | did it break twice in one day? | 21:10 |
| matty_dubs | NobodyCam: Heh, I over-did that earlier today! | 21:10 |
| jroll | oh | 21:11 |
| matty_dubs | NobodyCam: Some sort of keyboard issue or something caused me to have " if not (driver_info['virt_type'] == 'ontap_sp')))))))):" | 21:11 |
| jroll | devananda: I probably need to rebuild my venv, yeah? | 21:11 |
| NobodyCam | :O ieek | 21:11 |
| devananda | yes | 21:11 |
| jroll | :) | 21:11 |
| jroll | linggao: ^ | 21:11 |
| linggao | devananda, so it is merged already. | 21:11 |
| devananda | jroll, linggao: pip install -U -r requirements.txt -r test-requriements.txt | 21:11 |
| linggao | thanks jroll. | 21:11 |
| devananda | woops, typo, but you getthe idea | 21:11 |
| jroll | devananda: but it takes so long :( :P | 21:11 |
| devananda | jroll: that's faster | 21:12 |
| devananda | just update, dont rebuild the whole thing | 21:12 |
| jroll | I assumed the long part was pip | 21:12 |
| jroll | ohhh, right, it won't reinstall everythign | 21:12 |
| jroll | good call, thanks | 21:12 |
| *** eguz has quit IRC | 21:12 | |
| * jroll is not sure his brain is at 100% right now | 21:12 | |
| *** eghobo has joined #openstack-ironic | 21:13 | |
| dividehex | Does anyone know if the consistent hash prevents nodes from mapping to conductors not in the same availability zones? | 21:13 |
| *** jbjohnso has quit IRC | 21:13 | |
| devananda | dividehex: ironic conductors don't know anything about nova host aggregates | 21:13 |
| russell_h | dividehex: it doesn't | 21:13 |
| devananda | so, no | 21:14 |
| comstud | devananda: cool :) | 21:14 |
| devananda | dividehex: unless your AZs use separate DB's | 21:14 |
| dividehex | ok thanks | 21:14 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Better handling of missing drivers https://review.openstack.org/83572 | 21:14 |
| russell_h | I think we'll eventually wind up wanting to do something like that, but I don't think it'll be based on host aggregates | 21:15 |
| devananda | rack-aware scheduling :) | 21:16 |
| russell_h | if anything, you might base host aggregates on the same location info Ironic uses | 21:16 |
| *** matty_dubs is now known as matty_dubs|gone | 21:16 | |
| rloo | russell_h: didn't realize you are around. I just commented on your review 81919. Let me know what you think. | 21:16 |
| linggao | devananda, I did pip install -U -r requirements.txt -r test-requriements.txt, and still got ironic.conf.sample is not up to date. | 21:16 |
| russell_h | devananda: right, but not just scheduling of instances, but also of conductor mapping | 21:16 |
| devananda | linggao: sorry, i had a typo -- it should be "pip install -U -r test-requirements.txt" | 21:16 |
| *** eguz has joined #openstack-ironic | 21:16 | |
| russell_h | probably "scheduling of conductor mapping" makes no sense | 21:17 |
| devananda | russell_h: mapping based on proximity | 21:17 |
| russell_h | right | 21:17 |
| russell_h | people are going to want to do top of rack conductors or whatever | 21:17 |
| devananda | sure | 21:17 |
| russell_h | anyway, that seems like something that can be reasonably added later | 21:17 |
| devananda | but that creates a SPoF for each rack | 21:17 |
| devananda | and not having that SPoF is sorta the point ... | 21:17 |
| russell_h | right, you'd still want a fallback | 21:17 |
| russell_h | thats where it gets... more fun | 21:18 |
| devananda | so a preferred-conductor | 21:18 |
| russell_h | right | 21:18 |
| devananda | so then we have a different mapping function. hash-based and proximity-preference-based | 21:18 |
| linggao | devananda, same error. | 21:18 |
| russell_h | it could literally be that every node maps to the conductor at the top of its rack, with its secondary in the next rack over | 21:18 |
| devananda | linggao: hmm... what is the diff? | 21:18 |
| jroll | linggao: your connection to pypi must be much better than mine :) | 21:18 |
| russell_h | but making it suitably generic sounds hard | 21:18 |
| devananda | russell_h: we need rack discovery and such first, but yes, i agree that would /eventually/ be nice :) | 21:19 |
| russell_h | which is why I'm really happy to not worry about it for now | 21:19 |
| russell_h | right | 21:19 |
| linggao | let me try the diff... | 21:19 |
| russell_h | rloo: let me take a look | 21:19 |
| russell_h | rloo: oohh, thank you! | 21:19 |
| russell_h | I'll work through this | 21:20 |
| russell_h | I didn't realize which review that was at first | 21:20 |
| rloo | russell_h: that's the first review of many, right? :-) | 21:20 |
| *** eghobo has quit IRC | 21:21 | |
| *** eguz has quit IRC | 21:21 | |
| *** eghobo has joined #openstack-ironic | 21:22 | |
| russell_h | rloo: so, regarding your top-level comment, I thought some about that, but I like keeping them separate more | 21:22 |
| russell_h | fundamentally, I view per-node vendor passthru methods as totally different methods than per-driver ones | 21:22 |
| jroll | JoshNang: this maybe needs to be fixed? https://review.openstack.org/#/c/84795/25/etc/ironic/ironic.conf.sample | 21:22 |
| NobodyCam | seems odd that that pxeandSeamicroDriver is the only class that checks if it loaded :-p | 21:23 |
| russell_h | so making them the same actual method on the VendorInterface gets odd | 21:23 |
| jroll | JoshNang: just regen the sample config | 21:23 |
| JoshNang | heh that's what i did. | 21:23 |
| russell_h | rloo: in an ideal world, I would rename "vendor_passthru()" to "node_vendor_passthru()" | 21:23 |
| JoshNang | and gave me that. | 21:23 |
| russell_h | rloo: but thats a whole can of worms I don't want to get into | 21:23 |
| rloo | russell_h: Ok, so a vendor passthru method is specifically either geared towards the node, or not. | 21:23 |
| jroll | JoshNang: pip install -U -r requirements.txt -r test-requirements.txt | 21:23 |
| russell_h | right | 21:23 |
| jroll | JoshNang: those are changes in keystone | 21:24 |
| russell_h | rloo: the best example I have of this is what we're doing in the agent driver, where we need a way to look up what node an agent belongs to | 21:24 |
| JoshNang | ah awesome | 21:24 |
| rloo | russell_h: so if method1 is not node-specific, you cannot invoke it via nodes/.../vendor_passthru. | 21:24 |
| russell_h | correct | 21:24 |
| rloo | russell_h: and it is/will be obvious to whomever invokes the methods, to know what the method names are and whether to go via /nodes/.. or /drivers/... | 21:25 |
| russell_h | in the agent driver, we pass in a JSON document with MAC addresses and serial #s and whatnot to /drivers/.../vendor_passthru/lookup | 21:25 |
| russell_h | and the agent driver tries to find the node in the database | 21:25 |
| rloo | russell_h: ok. fair enough. | 21:25 |
| devananda | NobodyCam: yes, we should do the same for pyghmi, and PXE should check the presence of /tftpboot/ and tfptd and such during its __init__ too | 21:25 |
| NobodyCam | so devananda how many +2's does 83572 need. If I add mine that will make 4 | 21:25 |
| russell_h | and if it doesn't, it'll can optionally add it (not sure if that code is written yet) | 21:25 |
| NobodyCam | devananda: yes! | 21:26 |
| russell_h | rloo: but it wouldn't make sense to do like /nodes/.../vendor_passthru/lookup | 21:26 |
| devananda | NobodyCam: lol - that failed to merge, i just rebased it. if it passes jenkins, we should +A it | 21:26 |
| russell_h | thats how I'm thinking about it anyway | 21:26 |
| NobodyCam | lol ack will keep an eye on it and land it | 21:26 |
| *** Mikhail_D_ltp has quit IRC | 21:26 | |
| jroll | russell_h: wdyt about making both passthru things at /drivers/.../vendor_passthru/method, and adding ?node=uuid if it's for a node? | 21:27 |
| jroll | lucas mentioned this the other day | 21:27 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Use DiskPartitioner https://review.openstack.org/83399 | 21:27 |
| rloo | russell_h: but even with that, why at the driver level, do you need to separate into (node_)vendor_passthru() and driver_vendor_passthru(). Unless that makes it easier to check/distinguish the two. | 21:27 |
| linggao | devananda, here is the diff. http://paste.openstack.org/show/76174/. | 21:28 |
| linggao | devananda, something wrong with my keystone level again. It had the same problem but magically healled. | 21:28 |
| devananda | linggao: yes, the python-keystoneclient library in your environment needed to be updated | 21:29 |
| *** zdiN0bot has quit IRC | 21:29 | |
| jroll | devananda, linggao: JoshNang's patch that I mentioned above has the same diff :/ | 21:29 |
| devananda | linggao: that wil lsolve your problem | 21:29 |
| devananda | the ironic.conf.sample on the master branch is correct right now | 21:29 |
| rloo | russell_h. I don't see why you can't change vendor_passthru() to node_vendor_passthru() then. but that could be a future change I guess. | 21:29 |
| jroll | oh hmm | 21:29 |
| russell_h | rloo: right | 21:29 |
| devananda | so anyone getting that error has an out-of-date venv | 21:29 |
| JoshNang | just did pip -U... and generate but nothing changed | 21:29 |
| russell_h | rloo: I'd worry about anyone with an out-of-tree driver | 21:30 |
| JoshNang | i just rebuilt my whole vagrant box this morning (whoops) | 21:30 |
| russell_h | I don't know the policy on that, are driver interfaces meant to be stable? | 21:30 |
| jroll | devananda: what is the "correct" version of python-keystoneclient for this? | 21:30 |
| rloo | russell_h: How could someone have an out-of-tree driver but updated conductor/etc? | 21:30 |
| russell_h | but in any case, IMO it belongs in a separate change if it should happen at all | 21:30 |
| jroll | devananda: I wonder if maybe keystone did another release and now our sample is wrong again | 21:30 |
| devananda | jroll: i see 0.8.0 locally | 21:30 |
| russell_h | rloo: drivers just get loaded from your python path | 21:30 |
| jroll | devananda: yay version ranges in global requirements. | 21:30 |
| russell_h | so they can be versioned entirely independently of Ironic | 21:30 |
| devananda | yea... | 21:30 |
| russell_h | rloo: you just drop some info in your setup.cfg, install it somewhere on your python path, and it'll show up to Ironic as a driver | 21:31 |
| devananda | russell_h: ironic driver API is meant to be stable at certain points | 21:31 |
| russell_h | stevedore is actually kind of neat | 21:31 |
| linggao | mine is 0.7.1 | 21:31 |
| russell_h | right, thats sort of what I figured | 21:31 |
| rloo | russell_h: ah yes, that's true. Although I don't think anyone has been worried about that since eg there's some new ManagementInterface coming up. | 21:31 |
| russell_h | we could do a deprecation thing | 21:31 |
| jroll | linggao: pip install -U python-keystoneclient==0.8.0 | 21:32 |
| devananda | given that folks are writing out-of-tree drivers, we need to have deprecation plan / cycle for driver API changes | 21:32 |
| russell_h | rloo: so, I think (although I can't swear I verified this) that the ManagementInterface is optional | 21:32 |
| devananda | or just don't break it | 21:32 |
| russell_h | rloo: so if you're missing it, things will still work, although... if things get moved over into it... it might not :) | 21:33 |
| rloo | russell_h: I think it is optional, but I think some functions are going to move to that interface from whatever existing interface. Not sure though. | 21:33 |
| russell_h | right | 21:33 |
| devananda | so the purpose of vendor-passthru, IMO, is to allow methods to "incubate" there | 21:34 |
| JoshNang | installed python-keystoneclient==0.8.0, still same sample config :/ | 21:34 |
| rloo | devananda. can we add a note somewhere, that the driver API isn't yet stable or something like that? | 21:34 |
| jroll | JoshNang: o.o | 21:34 |
| devananda | if we decide taht a given method is critical to the functionality of all drivers, it should mvoe to a core interface | 21:34 |
| * jroll is *still* upgrading things | 21:34 | |
| russell_h | at risk of sounding really aggressive, and as someone running a basically out-of-tree driver, kind of hypocritical, Ironic is very young to get bogged down worrying too much about backwards compatibility with itself | 21:34 |
| devananda | rloo: what part of the core driver API is not stable? | 21:34 |
| devananda | JoshNang: is 14c61d69a20c9afaee8e534f0143874f751d1783 in your git history? | 21:35 |
| *** rloo has quit IRC | 21:35 | |
| jroll | devananda: it could be argued that vendor_passthru is not stable (ever) | 21:35 |
| devananda | jroll: precisely | 21:35 |
| jroll | or more undefined then unstable | 21:35 |
| *** rloo has joined #openstack-ironic | 21:35 | |
| rloo | devananda. Nothing yet. russell_h and I were discussing backwards compatibility if we change the driver API. | 21:35 |
| *** jrist has quit IRC | 21:36 | |
| JoshNang | devananda: "Update ironic.conf.sample" yup | 21:36 |
| devananda | vendor_passthru should never be considered stable -- it's whole purpose is to allow experimentation. perhaps we should rename it that way | 21:36 |
| jroll | rloo: are you saying we should document that it is not stable, and not worry about backwards compat? | 21:36 |
| devananda | (kidding on the rename) | 21:36 |
| *** jrist has joined #openstack-ironic | 21:36 | |
| russell_h | devananda: the last service I built has a "/__experimental" namespace :) | 21:36 |
| * NobodyCam was typing reply to rename comment | 21:36 | |
| linggao | devananda, jroll, installed 0.8.0, still has the same error. So I'l have to leave it as is since I am not changing the sample file anyway. :). It will work tomorrow just like last time. ;) | 21:36 |
| rloo | jroll: anything so we can move forward w/o having to worry ;) | 21:36 |
| linggao | thanks for the help | 21:36 |
| jroll | linggao: hm... | 21:37 |
| russell_h | devananda: which hasn't stopped anyone from depending on it... | 21:37 |
| openstackgerrit | linggao proposed a change to openstack/ironic: Handle API error for XML client https://review.openstack.org/88448 | 21:37 |
| russell_h | especially when they discovered the JSON Home API hidden in there | 21:37 |
| jroll | rloo: I'm ok with that, to be honest, but I like moving fast :) | 21:37 |
| devananda | russell_h: the alternative here is not to allow *anything* in the API until it's deemed stable | 21:38 |
| devananda | nova addressed this with API extensions | 21:38 |
| linggao | have to go. good night everyone. | 21:38 |
| rloo | devananda: this came up, cuz russell_h is adding a driver_vendor_passthru to VendorInterface, and I first thought why doesn't everything go through vendor_passthru(). But if we need to have two passthru's, then maybe vendor_passthru -> node_vendor_passthru. Dunno. | 21:38 |
| NobodyCam | night linggao | 21:38 |
| jroll | night ling | 21:38 |
| devananda | which ended up being a bit of a nightmare for them, since a) lots of extensions essentially becaome "core" functionalty, but stayed nominally extensions, and b) different deployers enabled different entensions | 21:38 |
| devananda | *extensions | 21:38 |
| devananda | leading to different functionality across nova implementations of the same version | 21:38 |
| *** linggao has quit IRC | 21:39 | |
| russell_h | devananda: I'm on your side here, you sold me on vendor_passthru a long time ago | 21:39 |
| devananda | russell_h: :) | 21:39 |
| russell_h | the particular implementation is a little ghetto, but basically fine | 21:39 |
| devananda | rloo, russell_h: by the way, did you see my reply on the BP? | 21:39 |
| russell_h | yep, replying to the reply now | 21:39 |
| russell_h | actually for like the last 2 hours | 21:40 |
| russell_h | keep getting distracted | 21:40 |
| * rloo looks now. | 21:40 | |
| devananda | rloo: i get the reason for a driver-vendor-passthru. i think the name is too long, but the functionality seems correct - -for their use case. i'm concerned it breaks other uses though | 21:40 |
| devananda | (also, names are hard) | 21:40 |
| *** rloo has quit IRC | 21:41 | |
| *** rloo has joined #openstack-ironic | 21:41 | |
| *** rloo has joined #openstack-ironic | 21:42 | |
| russell_h | devananda: replied to the BP | 21:43 |
| russell_h | now, with my name on it | 21:43 |
| devananda | :) | 21:45 |
| russell_h | blueprints are horrible | 21:45 |
| devananda | yep | 21:45 |
| *** vkozhukalov has quit IRC | 21:47 | |
| NobodyCam | devananda: lol... Discuss summit proposals Then do anything else, if we have time.... | 21:49 |
| jroll | uh oh | 21:50 |
| NobodyCam | should we add the bp repo / template idea to agenda? | 21:50 |
| * jroll meant to put in more summit proposals today | 21:50 | |
| devananda | NobodyCam: yes | 21:52 |
| devananda | jroll: do it. sudnay's the deadline | 21:52 |
| devananda | also, we have 14 proposals right now and ONLY FOUR SLOTS | 21:52 |
| devananda | so please, add more :) | 21:52 |
| devananda | (no, really) | 21:52 |
| russell_h | heh | 21:53 |
| adam_g | devananda, what was the bug you mentioned re resource tracking? | 21:53 |
| jroll | devananda: I know :| | 21:53 |
| jroll | I'll find time. somewhere. somehow. | 21:53 |
| *** newell has quit IRC | 21:55 | |
| *** foexle has joined #openstack-ironic | 21:56 | |
| *** Haomeng has quit IRC | 21:57 | |
| devananda | adam_g: i'm not finding it right now... | 21:57 |
| devananda | adam_g: IIRC, someone in -nova was trying to remove the host_manager classes since they aren't used any more, and then noticed that ironic seems to /think/ they're being used | 21:58 |
| adam_g | hmm | 21:58 |
| comstud | The host manager class is used | 21:59 |
| adam_g | it looks like exposing nodes /w no associated instance but powered on or in maitanence mode really screws with nova's resource tracking | 21:59 |
| comstud | someone would be on crack | 21:59 |
| devananda | comstud: i may be completely misstating the problem | 22:00 |
| comstud | I saw something maybe related to no-db schedulding | 22:00 |
| comstud | or a discussion based off of it | 22:00 |
| comstud | devananda: np, just confirming it's definitely used | 22:00 |
| comstud | :) | 22:00 |
| devananda | comstud: i believe their conjecture was that http://git.openstack.org/cgit/openstack/ironic/tree/ironic/nova/scheduler/ironic_host_manager.py#n56 is not used | 22:01 |
| comstud | oh, ironic in specific, k | 22:02 |
| devananda | comstud: ah. it's teh HostStates class | 22:03 |
| devananda | i dont see anywhere that it's used | 22:03 |
| devananda | in nova | 22:03 |
| devananda | gah | 22:04 |
| devananda | it helps if I don't have a typo in my grep line | 22:04 |
| devananda | ignore me -- it's totally used | 22:04 |
| comstud | :) | 22:04 |
| NobodyCam | devananda: wanta take a quick look at : https://review.openstack.org/#/c/88370 | 22:05 |
| rloo | devananda: I think it was this email about BaremetalHostManager http://lists.openstack.org/pipermail/openstack-dev/2014-April/031903.html | 22:06 |
| russell_h | rloo: I seem to type a lot of "i"s and "y"s out of order | 22:06 |
| rloo | russell_h: there weren't that many, were there? ;) | 22:06 |
| devananda | rloo: yes! that's it :) | 22:06 |
| devananda | adam_g: ^ | 22:06 |
| adam_g | ah | 22:07 |
| adam_g | okay thats a separate issue than what im looking at | 22:08 |
| rloo | btw russell_h, because I added a comment to your review, you should probably reply so others know. And when I'm back on Tues, I hope that it will have been approved ;) | 22:08 |
| adam_g | for ironic to expose nodes /w power_state ON or maintenance mode, nova will need a corresponding patch to its resource tracking | 22:09 |
| devananda | NobodyCam: I dont think validate() is going to set the power state | 22:09 |
| NobodyCam | humm | 22:10 |
| devananda | adam_g: nova.virt.ironic, or nova.compute.resource_tracker? | 22:10 |
| adam_g | devananda, nova.compute.resource_tracker | 22:10 |
| devananda | adam_g: that's not pluggable, which means we cant do that out of tree | 22:11 |
| devananda | time to get a patch merged to nova | 22:11 |
| adam_g | devananda, the dict ironic return for resources contains keys that get reset by _update_usagte_from_instances() | 22:11 |
| adam_g | http://git.openstack.org/cgit/openstack/nova/tree/nova/compute/resource_tracker.py#n596 | 22:12 |
| adam_g | attempting to calculate powered on or nodes in maitanence as part of the used resources from the driver does not good wrt actual resource tracking | 22:12 |
| NobodyCam | devananda: https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/ssh.py#L371-L372 | 22:13 |
| devananda | oh, hah | 22:13 |
| devananda | adam_g: so even though our HostManager is consuming all the resources, nova is overriding that? | 22:13 |
| NobodyCam | well | 22:13 |
| NobodyCam | that check the ssh connection | 22:14 |
| adam_g | devananda, looks that way, at least when there are no instances associated iwth the nodes | 22:14 |
| devananda | NobodyCam: right. it doesn't set the node.power_state property | 22:14 |
| devananda | NobodyCam: try just setting the power state there :) | 22:17 |
| devananda | NobodyCam: though if you need error detection, you should set-and-poll | 22:18 |
| NobodyCam | lol that would be easy.. lol.. face in palms :-p | 22:18 |
| devananda | ok, i need to run for a while... didn't take a break earlier when i should have. | 22:19 |
| devananda | bbiah | 22:19 |
| *** jgrimm has quit IRC | 22:20 | |
| *** dwalleck_ has quit IRC | 22:24 | |
| NobodyCam | I to am going to run | 22:25 |
| *** zdiN0bot has joined #openstack-ironic | 22:28 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 22:34 |
| jroll | JoshNang: I wouldn't say that method "isn't implemented", I would say there's just no validation done there. | 22:39 |
| jroll | but I guess it's fine | 22:39 |
| openstackgerrit | A change was merged to openstack/ironic: Better handling of missing drivers https://review.openstack.org/83572 | 22:39 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 22:40 |
| JoshNang | :) | 22:40 |
| jroll | gahhhhhhh | 22:40 |
| * jroll +1's again | 22:40 | |
| jroll | so many patchsets | 22:40 |
| JoshNang | only a couple more and i'll break 100 emails in that thread! | 22:41 |
| JoshNang | i wonder if gmail has a limit before it starts a new thread. | 22:41 |
| jroll | heh | 22:41 |
| *** derekh has joined #openstack-ironic | 22:41 | |
| jroll | it does, but I forget what it is. 150 maybe | 22:41 |
| *** rwsu has quit IRC | 22:44 | |
| *** jistr has quit IRC | 22:46 | |
| *** rwsu has joined #openstack-ironic | 22:48 | |
| *** zdiN0bot1 has joined #openstack-ironic | 22:49 | |
| *** zdiN0bot has quit IRC | 22:49 | |
| openstackgerrit | A change was merged to openstack/ironic: Refactor sync_power_states tests to not use DB https://review.openstack.org/87076 | 22:55 |
| openstackgerrit | A change was merged to openstack/ironic: Make sync_power_states yield https://review.openstack.org/84862 | 22:55 |
| *** zdiN0bot1 has quit IRC | 22:58 | |
| *** zdiN0bot has joined #openstack-ironic | 23:04 | |
| *** zdiN0bot has quit IRC | 23:04 | |
| *** zdiN0bot has joined #openstack-ironic | 23:06 | |
| *** zdiN0bot has quit IRC | 23:13 | |
| *** zdiN0bot has joined #openstack-ironic | 23:15 | |
| *** zdiN0bot has quit IRC | 23:17 | |
| *** zdiN0bot has joined #openstack-ironic | 23:21 | |
| *** eghobo has quit IRC | 23:25 | |
| *** zdiN0bot has quit IRC | 23:26 | |
| *** stevehuang has quit IRC | 23:38 | |
| devananda | back for a bit, then will be off to dinner | 23:56 |
| * devananda tests https://review.openstack.org/#/c/83399/ | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
