| *** eghobo has quit IRC | 00:07 | |
| *** eghobo has joined #openstack-ironic | 00:07 | |
| *** newell has quit IRC | 00:07 | |
| *** eguz has joined #openstack-ironic | 00:08 | |
| *** eghobo has quit IRC | 00:12 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 00:22 |
|---|---|---|
| *** eguz has quit IRC | 00:23 | |
| *** zdin0bot has joined #openstack-ironic | 00:35 | |
| *** tteggel_ has joined #openstack-ironic | 00:38 | |
| *** tteggel has quit IRC | 00:48 | |
| *** tteggel_ is now known as tteggel | 00:48 | |
| *** zdin0bot has quit IRC | 00:52 | |
| *** zdin0bot has joined #openstack-ironic | 00:52 | |
| *** zdin0bot has quit IRC | 00:53 | |
| *** zdin0bot has joined #openstack-ironic | 00:53 | |
| *** yongli has joined #openstack-ironic | 01:00 | |
| *** rloo has quit IRC | 01:08 | |
| *** nosnos has joined #openstack-ironic | 01:33 | |
| *** killer_prince has quit IRC | 01:42 | |
| *** eghobo has joined #openstack-ironic | 01:47 | |
| *** eghobo has quit IRC | 01:57 | |
| *** zdin0bot has quit IRC | 01:59 | |
| *** zdin0bot has joined #openstack-ironic | 02:00 | |
| *** early has quit IRC | 02:26 | |
| *** yfujioka has joined #openstack-ironic | 02:29 | |
| *** early has joined #openstack-ironic | 02:31 | |
| *** yfujioka has quit IRC | 02:34 | |
| *** zdin0bot has quit IRC | 02:35 | |
| *** yongli has quit IRC | 02:35 | |
| *** yongli has joined #openstack-ironic | 02:49 | |
| *** coolsvap has joined #openstack-ironic | 02:52 | |
| *** eghobo has joined #openstack-ironic | 02:56 | |
| *** eghobo has quit IRC | 03:09 | |
| *** zdin0bot has joined #openstack-ironic | 03:13 | |
| *** harlowja is now known as harlowja_away | 03:13 | |
| *** zdin0bot has quit IRC | 03:15 | |
| *** zdin0bot has joined #openstack-ironic | 03:33 | |
| *** flwang__ has quit IRC | 03:36 | |
| *** nosnos has quit IRC | 03:45 | |
| *** coolsvap has quit IRC | 03:54 | |
| *** eghobo has joined #openstack-ironic | 04:12 | |
| *** coolsvap has joined #openstack-ironic | 04:13 | |
| raies | greghaynes: | 04:15 |
| *** rameshg87 has joined #openstack-ironic | 04:45 | |
| *** lazy_prince has joined #openstack-ironic | 04:45 | |
| *** nosnos has joined #openstack-ironic | 04:53 | |
| *** matsuhashi has joined #openstack-ironic | 04:54 | |
| *** matsuhashi has quit IRC | 05:24 | |
| *** zdin0bot has quit IRC | 05:26 | |
| *** pradipta_away has quit IRC | 05:26 | |
| *** zdin0bot has joined #openstack-ironic | 05:26 | |
| *** sabah has joined #openstack-ironic | 05:28 | |
| *** zdin0bot has quit IRC | 05:31 | |
| *** matsuhashi has joined #openstack-ironic | 05:40 | |
| *** max_lobur has joined #openstack-ironic | 05:47 | |
| *** coolsvap1 has joined #openstack-ironic | 05:48 | |
| *** coolsvap has quit IRC | 05:48 | |
| *** coolsvap1 is now known as coolsvap|afk | 05:49 | |
| *** coolsvap|afk is now known as coolsvap | 05:50 | |
| *** sabah has quit IRC | 06:02 | |
| *** ifarkas has joined #openstack-ironic | 06:04 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/88508 | 06:07 |
| *** matsuhashi has quit IRC | 06:14 | |
| *** vkozhukalov has joined #openstack-ironic | 06:15 | |
| greghaynes | raies: HAI | 06:17 |
| *** matsuhashi has joined #openstack-ironic | 06:24 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 06:36 | |
| *** tatyana_ has joined #openstack-ironic | 06:49 | |
| *** tatyana has quit IRC | 06:51 | |
| *** tatyana_ is now known as tatyana | 06:51 | |
| *** rwsu has quit IRC | 06:57 | |
| *** foexle has joined #openstack-ironic | 06:59 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add IloPower module for IloDriver https://review.openstack.org/90294 | 07:00 |
| *** nosnos has quit IRC | 07:02 | |
| *** eghobo has quit IRC | 07:02 | |
| *** vigneshvar_ has joined #openstack-ironic | 07:03 | |
| *** viktors|afk is now known as viktors | 07:13 | |
| *** matsuhashi has quit IRC | 07:15 | |
| *** matsuhashi has joined #openstack-ironic | 07:15 | |
| *** yuriyz has joined #openstack-ironic | 07:16 | |
| *** ndipanov has joined #openstack-ironic | 07:26 | |
| *** matsuhashi has quit IRC | 07:28 | |
| *** matsuhashi has joined #openstack-ironic | 07:29 | |
| *** tatyana has quit IRC | 07:33 | |
| *** vkozhukalov has quit IRC | 07:37 | |
| *** sabah has joined #openstack-ironic | 07:50 | |
| *** martyntaylor has joined #openstack-ironic | 07:56 | |
| *** BadCub has quit IRC | 07:59 | |
| *** mkerrin has quit IRC | 07:59 | |
| *** matsuhashi has quit IRC | 08:06 | |
| *** mkerrin has joined #openstack-ironic | 08:06 | |
| *** matsuhashi has joined #openstack-ironic | 08:07 | |
| *** derekh has joined #openstack-ironic | 08:09 | |
| *** romcheg has quit IRC | 08:10 | |
| *** romcheg has joined #openstack-ironic | 08:10 | |
| *** nosnos has joined #openstack-ironic | 08:11 | |
| *** Mikhail_D_ltp has quit IRC | 08:13 | |
| *** foexle has quit IRC | 08:25 | |
| *** foexle has joined #openstack-ironic | 08:25 | |
| *** lsmola has joined #openstack-ironic | 08:29 | |
| *** dguerri is now known as _dguerri | 08:30 | |
| *** _dguerri is now known as dguerri | 08:30 | |
| *** lucasagomes has joined #openstack-ironic | 08:32 | |
| *** dguerri is now known as _dguerri | 08:35 | |
| *** _dguerri is now known as dguerri | 08:35 | |
| *** matsuhashi has quit IRC | 08:36 | |
| *** nosnos has quit IRC | 08:46 | |
| *** romcheg has quit IRC | 08:50 | |
| *** foexle_ has joined #openstack-ironic | 08:51 | |
| Mikhail_D_wk | G'morning Ironic :) | 08:51 |
| *** romcheg has joined #openstack-ironic | 08:53 | |
| *** foexle has quit IRC | 08:54 | |
| *** zdiN0bot has joined #openstack-ironic | 08:56 | |
| *** Alexei_987 has joined #openstack-ironic | 08:56 | |
| Mikhail_D_wk | comstud: Hi! I have a question concerning the removal of `@ objectify` from `dbapi` ... When do you plan to do this, of course, if you plan? And how do you evaluate the work in the chain patches `Get rid object model `dict` methods`? | 08:58 |
| comstud | Mikhail_D_wk: I'm slowly working on it when I can between other things | 09:00 |
| *** romcheg has quit IRC | 09:00 | |
| comstud | The next thing on my list is removing it from 'get_nodeinfo_list' | 09:00 |
| *** zdiN0bot has quit IRC | 09:00 | |
| comstud | But that requires implementing a NodeList object in objects/node.py | 09:00 |
| comstud | and some support into Node to partially fill a Node and allow lazy-loading of other attributes | 09:01 |
| comstud | I may be working on that tomorrow (my time tomorrow :) it's 2am right now.) | 09:01 |
| comstud | Mikhail_D_wk: I'm not sure I understand the 2nd question.. what do you mean by 'evaluate' ? | 09:02 |
| comstud | Are you asking how I'm labeling my patches (commit msgs).. or are you asking about the side effects it can cause? | 09:03 |
| comstud | If you mean the latter, nothing really changes.. because at the same time, I'm swapping out calls in the code: dbapi.get_node() becomes Node.get(), etc... still the code still gets a Node object. | 09:04 |
| *** vkozhukalov has joined #openstack-ironic | 09:04 | |
| comstud | (in the case of get_nodeinfo_list, I'll be replacing dbapi.get_nodeinfo_list calls with: NodeList.get(attributes=attributes, filters=filters), etc.. | 09:05 |
| comstud | Not sure if that answered your question :) | 09:05 |
| Mikhail_D_wk | comstud: If you seen them, in the right direction I did these patches :) | 09:06 |
| comstud | The object model ones? | 09:06 |
| comstud | I saw.. I left a comment on one of them in the past | 09:07 |
| *** romcheg has joined #openstack-ironic | 09:07 | |
| comstud | The tests you're changing under ironic/tests/db don't actually need to change... | 09:07 |
| comstud | But it's okay if they do. Sqlalchemy supports both dict and attribute notation | 09:07 |
| Mikhail_D_wk | I'm sorry to answer for so long | 09:07 |
| comstud | (Just like our Node object does ATM) | 09:07 |
| comstud | no problem | 09:08 |
| Mikhail_D_wk | :) | 09:08 |
| comstud | I think I would skip modifying the test of ironic/tests/db/* | 09:08 |
| comstud | My work is going to make them return SQLAlchemy models or dicts | 09:08 |
| comstud | But... | 09:08 |
| comstud | The rest of our tests should really use .<field_name> notation, as you've been fixing | 09:09 |
| comstud | err | 09:09 |
| *** yongli has quit IRC | 09:09 | |
| comstud | I think I would skip modifying the rest* of ironic/tests/db/* | 09:09 |
| comstud | An example of my change: | 09:10 |
| comstud | get_nodeinfo_list returns a list of [object.Node] right now | 09:10 |
| *** jistr has joined #openstack-ironic | 09:10 | |
| comstud | er | 09:10 |
| comstud | that's a bad a example | 09:10 |
| comstud | it doesn't do that | 09:10 |
| comstud | hehe | 09:10 |
| Mikhail_D_wk | )) | 09:10 |
| comstud | I changed get_node() from returning a object.Node.. to returning the SQLAlchemy model | 09:11 |
| comstud | Changing the tests doesn't break anything because you can SQLAlchemyModel['foo'] OR SQLALchemyModel.foo | 09:11 |
| comstud | It's possible that I'll be making DBAPI return dicts, however | 09:11 |
| comstud | which is why I would stop converting *those* tests | 09:11 |
| Mikhail_D_wk | yes, I saw this | 09:12 |
| comstud | ok | 09:12 |
| comstud | My goal is to not have any use of 'dbapi' other than in objects/*.py | 09:12 |
| comstud | (there may be a few edge cases) | 09:12 |
| comstud | Did I answer your questions? | 09:13 |
| Mikhail_D_wk | clear :) thanks for explanation :) | 09:14 |
| Mikhail_D_wk | yeah :) | 09:14 |
| Mikhail_D_wk | comstud: And you said you already 2am? | 09:15 |
| comstud | yeah, 2am | 09:15 |
| comstud | time for bed, I think :) | 09:16 |
| Mikhail_D_wk | Then good night! :) | 09:16 |
| comstud | thanks! laters.. | 09:16 |
| * comstud & | 09:16 | |
| Mikhail_D_wk | Good thinking! :) | 09:16 |
| lucasagomes | comstud, ping re task, node | 09:18 |
| lucasagomes | comstud, https://github.com/openstack/ironic/blob/master/ironic/conductor/manager.py#L710 | 09:19 |
| lucasagomes | I think I would need a big clean up before that | 09:19 |
| lucasagomes | some other part of the code assumes that node is being passed as well | 09:19 |
| *** athomas has joined #openstack-ironic | 09:20 | |
| Mikhail_D_wk | Hi, lucasagomes :) | 09:21 |
| lucasagomes | Mikhail_D_wk, morning :D | 09:22 |
| *** radsy has joined #openstack-ironic | 09:27 | |
| *** radsy has joined #openstack-ironic | 09:27 | |
| *** rameshg87 has quit IRC | 09:42 | |
| *** vigneshvar_ has quit IRC | 09:44 | |
| *** matsuhashi has joined #openstack-ironic | 09:47 | |
| *** rameshg87 has joined #openstack-ironic | 09:49 | |
| *** radsy has quit IRC | 09:49 | |
| *** nosnos_ has joined #openstack-ironic | 09:52 | |
| *** zdiN0bot has joined #openstack-ironic | 09:56 | |
| *** zdiN0bot has quit IRC | 10:00 | |
| *** matsuhashi has quit IRC | 10:11 | |
| *** matsuhas_ has joined #openstack-ironic | 10:14 | |
| *** athomas has quit IRC | 10:17 | |
| *** athomas has joined #openstack-ironic | 10:23 | |
| *** matsuhas_ has quit IRC | 10:30 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add IloDriver and its utils https://review.openstack.org/89500 | 10:32 |
| *** matsuhashi has joined #openstack-ironic | 10:42 | |
| *** matsuhashi has quit IRC | 10:43 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add IloPower module for IloDriver https://review.openstack.org/90294 | 10:53 |
| *** rameshg87 has left #openstack-ironic | 10:53 | |
| *** dguerri is now known as _dguerri | 10:57 | |
| *** subah has joined #openstack-ironic | 10:59 | |
| *** sabah has quit IRC | 10:59 | |
| *** nosnos_ has quit IRC | 11:03 | |
| *** subah has quit IRC | 11:04 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic-python-agent: Updated from global requirements https://review.openstack.org/88722 | 11:16 |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/python-ironicclient: Updated from global requirements https://review.openstack.org/89244 | 11:23 |
| *** coolsvap is now known as coolsvap|afk | 11:29 | |
| *** romcheg has quit IRC | 11:45 | |
| *** lucasagomes is now known as lucas-hungry | 11:48 | |
| *** jdob has joined #openstack-ironic | 12:30 | |
| *** lazy_prince has quit IRC | 12:36 | |
| *** lucas-hungry is now known as lucasagomes | 12:56 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Implement more robust caching for master images https://review.openstack.org/85387 | 12:59 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add ManagementInterface https://review.openstack.org/86063 | 13:01 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the VendorPassthru interface https://review.openstack.org/90363 | 13:09 |
| *** linggao has joined #openstack-ironic | 13:14 | |
| *** romcheg has joined #openstack-ironic | 13:26 | |
| *** mkerrin has quit IRC | 13:28 | |
| *** zdin0bot has joined #openstack-ironic | 13:33 | |
| NobodyCam | good morning ironic, says the man making coffee | 13:33 |
| NobodyCam | TGIF! | 13:33 |
| zdin0bot | NobodyCam: morning :-) | 13:33 |
| NobodyCam | morning zdin0bot | 13:34 |
| NobodyCam | lucasagomes: Did you see my ping yesterday? | 13:35 |
| NobodyCam | *CORES* review jam monday 8:00am <- | 13:35 |
| lucasagomes | NobodyCam, morning | 13:36 |
| lucasagomes | NobodyCam, oh ah sure | 13:36 |
| lucasagomes | works for me | 13:36 |
| romcheg | Morning all! | 13:39 |
| romcheg | devananda: Congrats! | 13:40 |
| lucasagomes | NobodyCam, btw, did you test the oslo.messaging? | 13:40 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Remove 'node' parameter from the VendorPassthru interface https://review.openstack.org/90363 | 13:41 |
| anteaya | congratulations devananda | 13:42 |
| lucasagomes | morning romcheg anteaya | 13:42 |
| *** jbjohnso has joined #openstack-ironic | 13:42 | |
| anteaya | lucasagomes: morning | 13:43 |
| anteaya | I forgot what it felt like to have breakfast with the family | 13:43 |
| anteaya | :D | 13:43 |
| anteaya | morning romcheg | 13:43 |
| anteaya | morning NobodyCam | 13:43 |
| *** mdenny has joined #openstack-ironic | 13:48 | |
| NobodyCam | lucasagomes: I got my test env working late last night .. may test failed ... :( and it was to late for me to dig in to why | 13:50 |
| NobodyCam | looking now | 13:50 |
| lucasagomes | NobodyCam, ah no hurry, just to know | 13:51 |
| lucasagomes | thanks for testing it out :) | 13:51 |
| NobodyCam | morning anteaya :) | 13:51 |
| NobodyCam | :) we've missed you | 13:52 |
| NobodyCam | :-p | 13:52 |
| *** max_lobur has quit IRC | 13:52 | |
| anteaya | I've missed you too | 13:52 |
| NobodyCam | *(BLUSH)* | 13:52 |
| anteaya | always glad to come home | 13:52 |
| anteaya | :D | 13:52 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: set_console_mode to not overwrite node.last_error https://review.openstack.org/90376 | 13:56 |
| *** martyntaylor has quit IRC | 14:00 | |
| *** Shrews has quit IRC | 14:03 | |
| *** martyntaylor has joined #openstack-ironic | 14:04 | |
| *** Shrews has joined #openstack-ironic | 14:04 | |
| *** dtantsur has quit IRC | 14:06 | |
| comstud | lucasagomes: ah, ok. i think maybe we clean up task and node args later then. don't want to hold up the feature | 14:10 |
| comstud | although i see you put up a patch | 14:11 |
| lucasagomes | comstud, right, I submitted a new for the first patch of the series | 14:11 |
| comstud | ok | 14:11 |
| *** viktors has quit IRC | 14:11 | |
| lucasagomes | want tot see the feedback of that one before changing the others | 14:11 |
| comstud | yeah | 14:11 |
| lucasagomes | comstud, can u take a quick look? | 14:11 |
| comstud | reviewing now | 14:11 |
| comstud | yesp | 14:11 |
| lucasagomes | comstud, I think the db thing maybe should go to another patch, it's not really related to those changes | 14:12 |
| lucasagomes | that's better to be refactored at once for each module, instead partially do it on this patch | 14:12 |
| comstud | hehe, your patches are going to conflict with mine | 14:12 |
| comstud | I'll finish reviewing yours, if you look at one of mine? :) | 14:12 |
| lucasagomes | comstud, sure | 14:12 |
| comstud | https://review.openstack.org/#/c/89328/ | 14:12 |
| NobodyCam | lucasagomes: wow - http://paste.openstack.org/show/Jlw0gfSuy40v5GPdpE77/ | 14:16 |
| NobodyCam | not sure thats anthing from your patch... retaeting | 14:16 |
| NobodyCam | retesting even :-p | 14:16 |
| comstud | hrm | 14:17 |
| comstud | that's generally bad | 14:18 |
| *** jgrimm has joined #openstack-ironic | 14:19 | |
| comstud | lucasagomes: https://review.openstack.org/#/c/90363/2/ironic/tests/drivers/test_ipminative.py,unified | 14:24 |
| comstud | lucasagomes: were these tests passing the wrong arguments before?! | 14:24 |
| yuriyz | morning/evening Ironic | 14:24 |
| lucasagomes | comstud, they were... | 14:24 |
| lucasagomes | :( | 14:24 |
| comstud | nice | 14:24 |
| lucasagomes | they were passing node instead of task | 14:24 |
| lucasagomes | NobodyCam, looking | 14:24 |
| comstud | wel, only node. not task and node | 14:24 |
| comstud | hehe ok | 14:25 |
| yuriyz | lucasagomes, why code 202 here https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L96 | 14:25 |
| comstud | seems like those aren't tested quite right if it passed | 14:25 |
| lucasagomes | yuriyz, yeah that's async, shouldn't be 202 :( I will fix | 14:25 |
| lucasagomes | yuriyz, didn't have time to reply on the review, thanks for that | 14:26 |
| lucasagomes | comstud, all the spawn_after is getting _spawn_worker as it first parameter | 14:26 |
| lucasagomes | comstud, should we try to move this _spawn_worker logic to the taskmanager? | 14:27 |
| comstud | lucasagomes: deva wanted to keep it in the Manager | 14:28 |
| comstud | I think I agree with it being in the Manager as well | 14:28 |
| comstud | this way.. you could still use the task_manager.spawn_worker with say.. 'eventlet.spawn' | 14:29 |
| lucasagomes | yeah for me it looks a bit out of the place as-is | 14:29 |
| lucasagomes | yeah | 14:29 |
| comstud | in case you want to spawn but not use the worker pool | 14:29 |
| lucasagomes | the manager is the resposable for starting the async job, so it should have control over the thread_pool as well | 14:29 |
| lucasagomes | yeha | 14:29 |
| comstud | yeah | 14:29 |
| lucasagomes | I will add a comment about it, cause at least for me it feels a bit out of place to have the conductormanager controlling the pool, but the task being started by the taskmanager | 14:30 |
| comstud | well, the taskmanager code is generic | 14:31 |
| comstud | it doesn't care if you use the pool or not.. i kinda like it that way | 14:31 |
| comstud | but yeah, leave your opinion | 14:31 |
| comstud | in the review | 14:31 |
| lucasagomes | ack | 14:31 |
| anteaya | NobodyCam: you're awesome | 14:32 |
| NobodyCam | ???? | 14:32 |
| comstud | I can move it if there's consensus.. but this is what deva and i discussed | 14:32 |
| NobodyCam | oh the email :-0 | 14:32 |
| lucasagomes | comstud, right, the patch lgtm I will give it a try here | 14:32 |
| comstud | lucasagomes: you can think of this in the reverse... the worker pool right now doesn't require that you use a task. there could be a need to use the worker pool without having a task | 14:33 |
| comstud | lucasagomes: in which case, the worker pool should stay in the manager | 14:33 |
| comstud | hehe | 14:33 |
| lucasagomes | hmm | 14:33 |
| lucasagomes | right yeah I can see the benefits | 14:33 |
| lucasagomes | esp the flexibility of passing spawn_after(eventlet.spawn... as u mentioned there | 14:34 |
| comstud | yeah | 14:34 |
| comstud | there could be a case in the future where we always want to spawn a thread to finish a task, bypassing the pool | 14:34 |
| comstud | i'm not sure what case that would be :) but who knows | 14:34 |
| lucasagomes | heh yeah... well it sounds like some premature optimization... but I'm not totaly against it in this case | 14:35 |
| comstud | nod | 14:35 |
| comstud | i think the manager maybe wanting to use the worker pool without a task is more of a reason to keep it in the manager | 14:36 |
| comstud | even if we don't have a case for that right now, either. | 14:36 |
| lucasagomes | +2 | 14:36 |
| lucasagomes | +1 | 14:36 |
| comstud | :) | 14:36 |
| lucasagomes | I think I like it more after you explained the reasons why the pool should stay in the manager | 14:36 |
| comstud | i can think of a use case immediately for that | 14:37 |
| comstud | which is periodic tasks | 14:37 |
| comstud | for example, right now... | 14:37 |
| comstud | a couple of periodic tasks loop through a bunch of Nodes | 14:37 |
| comstud | lock and spawn workers | 14:37 |
| comstud | A limit of how many workers it uses was added to 1 of them.. | 14:38 |
| comstud | er, I guess only one of them does this (_check_deploy_timeouts) | 14:38 |
| comstud | but.. | 14:38 |
| comstud | it might be better if that periodic task just stole 1 worker | 14:38 |
| comstud | and did all of the locking of each node within that worker | 14:38 |
| comstud | instead of this weird hack to try to prevent it from eating 'x' workers | 14:39 |
| comstud | but anyway.. that's a separate issue | 14:39 |
| comstud | but if we changed that, we would end up spawning a worker with no task | 14:39 |
| lucasagomes | comstud, heh yeah, having a limit for the number of workers that can be spawned by a periodic task is a bit hacky indeed | 14:43 |
| lucasagomes | anyhoo, I'm going to give ur patch a go here | 14:43 |
| comstud | i'm going to propose a patch for that shortly. :) | 14:43 |
| comstud | working on it now.. kinda bugs me | 14:43 |
| lucasagomes | heh | 14:43 |
| lucasagomes | great | 14:43 |
| comstud | we'll see what ppl think in the review | 14:43 |
| lucasagomes | comstud, about that vendor_passthru... as part of the management interface series I'm removing all that vendorpasstrhu stuff from the ipmitool and ipmitnative drivers | 14:45 |
| comstud | ah right | 14:45 |
| lucasagomes | so my patches conflicts heh yay | 14:46 |
| lucasagomes | I didn't want to add more patches to that management interface series, need to get some merged firtst | 14:46 |
| lucasagomes | the linear dependency is getting really long | 14:46 |
| comstud | yeah i saw | 14:46 |
| comstud | i was working backwards thinking it was going to be small | 14:46 |
| comstud | until I got like 5 patches back | 14:46 |
| comstud | hehe | 14:46 |
| lucasagomes | yeah | 14:47 |
| NobodyCam | we need to land those | 14:47 |
| lucasagomes | NobodyCam, +10000000 | 14:47 |
| lucasagomes | hah | 14:47 |
| NobodyCam | how many cores do we have on line atm? | 14:47 |
| lucasagomes | NobodyCam, I put another patch up today for the first of the serie... will wait reviews on that to start fixing the others | 14:48 |
| NobodyCam | lol | 14:48 |
| lucasagomes | NobodyCam, 3 I think? me u and romcheg | 14:48 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Consider Glance checksum when caching master images https://review.openstack.org/90390 | 14:49 |
| NobodyCam | lucasagomes: hummm getting stranger and strnger | 14:52 |
| NobodyCam | http://paste.openstack.org/show/FqGf0GLZDcYh5Ki5B6mj/ | 14:52 |
| lucasagomes | NobodyCam, ouch | 14:52 |
| lucasagomes | NobodyCam, doesn't seems related to the oslo.messaging but hmmm | 14:52 |
| lucasagomes | weird anyway | 14:52 |
| NobodyCam | 2014-04-25 14:48:49.494 4221 ERROR wsme.api [-] Server-side error: "Node a4a4b637-a58d-4496-bf0b-b42f5d02b97d is locked by host undercloud-undercloud-khowqq7mhuco.novalocal, please retry after the cur | 14:54 |
| NobodyCam | rent operation is completed. | 14:54 |
| NobodyCam | we have to do something about the random nature of this failure | 14:54 |
| lucasagomes | yeah, we should retry those | 14:58 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 2 https://review.openstack.org/73256 | 14:58 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 3 https://review.openstack.org/86536 | 14:58 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Cleanup mock patch without `with` part 1 https://review.openstack.org/73223 | 14:58 |
| openstackgerrit | Mikhail Durnosvistov proposed a change to openstack/ironic: Get rid of the newline "\" https://review.openstack.org/66793 | 14:58 |
| NobodyCam | our gate jobs will never be stable | 14:58 |
| NobodyCam | :( | 14:58 |
| NobodyCam | bbt... brb | 14:59 |
| *** zdiN0bot1 has joined #openstack-ironic | 15:00 | |
| *** zdiN0bot1 has quit IRC | 15:04 | |
| *** foexle_ has quit IRC | 15:06 | |
| lucasagomes | :/ | 15:08 |
| *** zdin0bot has quit IRC | 15:08 | |
| *** zdin0bot has joined #openstack-ironic | 15:08 | |
| *** dtantsur has joined #openstack-ironic | 15:11 | |
| jroll | devananda: congrats :) | 15:12 |
| jroll | happy friday ironic :) | 15:12 |
| *** BadCub has joined #openstack-ironic | 15:14 | |
| *** stephenpearson has joined #openstack-ironic | 15:16 | |
| NobodyCam | happy friday jroll | 15:16 |
| NobodyCam | lucasagomes: when you tested the messaging patch did you attempt to spin up more then one node? | 15:26 |
| lucasagomes | jroll, happy friday | 15:26 |
| lucasagomes | NobodyCam, hmmm not that I remember | 15:26 |
| lucasagomes | NobodyCam, I can try to give it a go | 15:26 |
| NobodyCam | I ask because I in my testing seed deploys undercloud but under cloud is not deploying overcloud | 15:26 |
| NobodyCam | i am going to retest with out hte patch | 15:27 |
| lucasagomes | right | 15:27 |
| lucasagomes | lemme give it a try as well | 15:27 |
| *** newell has joined #openstack-ironic | 15:29 | |
| NobodyCam | need to rebuild images for the test... will take a few min | 15:29 |
| *** kathay has joined #openstack-ironic | 15:31 | |
| *** martyntaylor has quit IRC | 15:40 | |
| *** martyntaylor has joined #openstack-ironic | 15:43 | |
| *** shakamunyi has joined #openstack-ironic | 15:44 | |
| agordeev | jroll: JayF: russell_h: morning :) wdyt about hw discovery improvement? -> https://etherpad.openstack.org/p/ipa-discovery-ext | 15:45 |
| NobodyCam | morning agordeev :) happy friday | 15:46 |
| agordeev | NobodyCam: morning! happy friday to you too | 15:46 |
| agordeev | also morning for everyone else | 15:47 |
| agordeev | have a nice TGIF | 15:47 |
| Mikhail_D_wk | good morning/afternoon/evening all! :) | 15:47 |
| agordeev | Mikhail_D_wk: morning | 15:47 |
| NobodyCam | morning Mikhail_D_wk | 15:47 |
| Shrews | Has anyone asked to have Ironic added to the subscription topic categories for the openstack-dev ML? | 15:48 |
| *** martyntaylor has left #openstack-ironic | 15:48 | |
| Shrews | guess not :) | 15:50 |
| NobodyCam | :-p | 15:50 |
| openstackgerrit | Russell Haering proposed a change to openstack/ironic: Drivers may expose a top-level passthru API https://review.openstack.org/81919 | 15:52 |
| lucasagomes | NobodyCam, deployed two machines http://paste.openstack.org/show/77105/ | 15:55 |
| NobodyCam | w00t.. did you deploy them at the same time. or do one then then other | 15:56 |
| lucasagomes | NobodyCam, one after the other | 15:57 |
| NobodyCam | ack | 15:57 |
| * kathay looks at NobodyCam | 16:00 | |
| kathay | syn-ack | 16:00 |
| *** zdiN0bot1 has joined #openstack-ironic | 16:00 | |
| NobodyCam | :-p | 16:01 |
| *** stephenpearson has quit IRC | 16:03 | |
| *** stephenpearson has joined #openstack-ironic | 16:03 | |
| *** eghobo has joined #openstack-ironic | 16:05 | |
| *** zdiN0bot1 has quit IRC | 16:05 | |
| NobodyCam | morning kathay :-p | 16:08 |
| devananda | g'morning, all | 16:08 |
| NobodyCam | good morning devananda :) | 16:09 |
| *** matty_dubs is now known as matty_dubs|lunch | 16:20 | |
| lucasagomes | morning devananda | 16:28 |
| lucasagomes | NobodyCam, devananda when you get some time, can you take a look at https://review.openstack.org/#/c/86063/ ? | 16:29 |
| lucasagomes | I fixed somethings that comstud pointed out, but I'd like ur opnion about it | 16:30 |
| lucasagomes | whether I should not pass the 'node' parameter for those methods | 16:30 |
| comstud | ya, I'd like to clean up where we pass both 'task' and 'node', since there's no reason to send node separately (it's in the task) | 16:31 |
| devananda | exactly | 16:31 |
| NobodyCam | I thought we tackled that already | 16:32 |
| *** athomas has quit IRC | 16:32 | |
| comstud | no, it's all over the place | 16:32 |
| comstud | :) | 16:32 |
| comstud | lucas started this: https://review.openstack.org/#/c/90363/ based on my feedback | 16:32 |
| devananda | lucasagomes: I have a concern adding NotSupportedByDriver | 16:33 |
| comstud | but there's a number of other cases after that still | 16:33 |
| lucasagomes | devananda, right | 16:33 |
| devananda | lucasagomes: if drivers' functionality differ enough within core/common interfaces that we need to add that, we're exposing an implementation detail to users that should be abstracted | 16:33 |
| devananda | we shouldn't be doing that | 16:34 |
| lucasagomes | devananda, hmm, problem is that some drivers implement get current bootdev, some other doesn't | 16:35 |
| lucasagomes | so I wanted to have a specific errors for those who doesn't support it | 16:35 |
| devananda | lucasagomes: that exc isn't used in this patch (probably used in follow on..) | 16:35 |
| comstud | $ egrep -R 'task, (node|task\.node)' ironic | wc -l | 16:35 |
| comstud | 138 | 16:35 |
| devananda | lucasagomes: is the intent for it to indicate that an interface isnt supported, or that a given method of an interface isn't supported? | 16:35 |
| comstud | heh | 16:35 |
| lucasagomes | devananda, yeah it was being used in the Fake driver, but I think I changed it after rloo review | 16:35 |
| JoshNang | agordeev: i like the improved data, which could be a new hardware manager. the driver saves whatever you send it to the DB. as far as what level of hardware discovery, i'm not sure i see much reason to add it. | 16:35 |
| lucasagomes | devananda, a given method of the interface isnt supported | 16:36 |
| devananda | lucasagomes: right. i think that violates the interface contract | 16:36 |
| devananda | lucasagomes: if a driver implements a core or common interface, it must implement all of it | 16:36 |
| lucasagomes | devananda, they all support setting the device, but not getting | 16:36 |
| devananda | lucasagomes: and we have invalidparameter for eg. "that boot device isn't valid on this hardware" | 16:37 |
| JoshNang | agordeev: basically, you add all your hardware managers to the config file, build the agent, and whatevers configured in hardware.get_manager() sends it data as part of its initial checkin, it gets saved to the db, yay! | 16:37 |
| lucasagomes | devananda, hmm | 16:37 |
| devananda | lucasagomes: what doesn't support getting it? | 16:37 |
| lucasagomes | devananda, https://github.com/stackforge/pyghmi/blob/master/pyghmi/ipmi/command.py#L125-L126 | 16:37 |
| *** yuriyz has quit IRC | 16:38 | |
| lucasagomes | ipmitool also I don't think it has a method to actually get the current bootdev | 16:38 |
| devananda | heh, ok | 16:39 |
| devananda | jbjohnso: comments ^ ? | 16:39 |
| devananda | lucasagomes: returning NULL or '' would seem better than an exception there | 16:39 |
| lucasagomes | devananda, right, we can have a boot_devices.NONE = None | 16:40 |
| lucasagomes | and I can return that in case I can't get it | 16:40 |
| devananda | hmm | 16:40 |
| jroll | agordeev: gah, forgot to look yesterday. will definitely try to get you feedback soon | 16:41 |
| devananda | at least that's my initial reaction. it's not not-implemented. if the hardware lacks support, that's not the driver's fault. | 16:41 |
| lucasagomes | devananda, would be better to return something like 'uknown' | 16:42 |
| lucasagomes | ? | 16:42 |
| devananda | lucasagomes: probably | 16:42 |
| devananda | in mysql, that's what NULL/NONE means :) | 16:42 |
| lucasagomes | devananda, right, ipmitool has a None options for boot dev (since we are basing our options in the ipmitool ones) | 16:43 |
| devananda | right | 16:44 |
| devananda | brb | 16:44 |
| *** athomas has joined #openstack-ironic | 16:45 | |
| lucasagomes | devananda, ack will make some changes in the patch then, cheers for the feedback | 16:45 |
| *** harlowja_away is now known as harlowja | 16:45 | |
| *** derekh has quit IRC | 16:47 | |
| *** mkerrin has joined #openstack-ironic | 16:48 | |
| *** igor has joined #openstack-ironic | 16:59 | |
| *** ndipanov has quit IRC | 16:59 | |
| lucasagomes | comstud, 138 :( | 17:00 |
| *** igor__ has quit IRC | 17:00 | |
| lucasagomes | comstud, yeah it's all over the place, the problem is also that some helper functions get boths methods as well, e.g https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe.py#L472 | 17:01 |
| *** zdiN0bot1 has joined #openstack-ironic | 17:01 | |
| comstud | lucasagomes: yeah | 17:04 |
| lucasagomes | get both paramters* | 17:05 |
| devananda | back | 17:05 |
| *** zdiN0bot1 has quit IRC | 17:06 | |
| *** eghobo has quit IRC | 17:07 | |
| *** openstackstatus has quit IRC | 17:07 | |
| *** eghobo has joined #openstack-ironic | 17:08 | |
| *** openstackstatus has joined #openstack-ironic | 17:08 | |
| NobodyCam | wb | 17:10 |
| *** stephenpearson has quit IRC | 17:10 | |
| kathay | good morning, NobodyCam :) | 17:15 |
| NobodyCam | :) | 17:15 |
| comstud | lucasagomes, devananda: so | 17:16 |
| comstud | lucasagomes, devananda: I guess a lot of these calls need to keep 'node' because it's possible you have locked multiple nodes | 17:16 |
| comstud | and you don't know which one in the task you're dealing with | 17:16 |
| lucasagomes | comstud, I gotta look deeper into it, but I dunno about any of those calls that are multinode | 17:17 |
| comstud | well | 17:17 |
| comstud | the problem is the helper methods should probably not assume | 17:17 |
| comstud | although I kinda dislike the multiple-node-locking in the first place | 17:18 |
| comstud | hehe | 17:18 |
| lucasagomes | yeah | 17:18 |
| lucasagomes | if the third one failed to get the lock should u unlock the others ? etc... | 17:18 |
| comstud | nod, right now must succeed locking.. others are unlocked and exception re-raised if not | 17:19 |
| comstud | er, all must succeed | 17:19 |
| comstud | but, i'm not sure this is used anywhere | 17:19 |
| devananda | right. well, planned feature that we moved away from and all | 17:19 |
| devananda | original intent of the task manager was to support synchronizing operations on >1 node, but we've pushed that logic much higher up the stack | 17:20 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add ManagementInterface https://review.openstack.org/86063 | 17:20 |
| comstud | nod | 17:20 |
| comstud | So we can remove multiple node support in task manager and clean this all up? | 17:20 |
| devananda | at this point, i don't think we have anything functionally operationg on >1 node at a time | 17:21 |
| comstud | I think that's required in order for us to fully ditch passing both 'task' and 'node' | 17:21 |
| devananda | so yes, it can probably be cleaned up... | 17:21 |
| devananda | though | 17:21 |
| lucasagomes | :) | 17:21 |
| devananda | there are some (specialized?) hardware types where it may still be handy | 17:21 |
| devananda | eg, dense compute with 4 SOCs on a single blade/cartridge | 17:21 |
| devananda | certain components are shared between them | 17:22 |
| devananda | so eg, flashing firmware on one node will affect the other 3 | 17:22 |
| lucasagomes | devananda, like operations that is done to one of the nodes will affect the others shared? | 17:22 |
| lucasagomes | hmm | 17:22 |
| devananda | yes | 17:22 |
| comstud | you'll need some decent changes I think to support that | 17:22 |
| devananda | right | 17:22 |
| comstud | essentially any operation would have to for node in task.resources: | 17:22 |
| comstud | call_method(task, node) | 17:22 |
| comstud | wherever you need | 17:22 |
| devananda | exactly. that was the original intent I had with task | 17:23 |
| lucasagomes | right, maybe leaving it fells into the premature optimization thing? since we would need many other parts to change to actually get those working | 17:23 |
| devananda | so we should still rip out node as a separate param :) | 17:23 |
| comstud | might be more simple to just wrap TaskManager with a TasksManager when you need it :) | 17:23 |
| comstud | list of tasks vs task with list of nodes | 17:23 |
| openstackgerrit | David Shrewsbury proposed a change to openstack/ironic: Implement instance rebuild in nova.virt.driver https://review.openstack.org/90429 | 17:24 |
| NobodyCam | grrrrrrr... Node cd950b2f-c437-4fd5-bc68-560726aca7d6 is locked by host undercloud-undercloud-iztcbftbhsul.novalocal, please retry after the current operation is completed. | 17:24 |
| NobodyCam | :-p | 17:24 |
| comstud | anyway, ok, cool | 17:24 |
| lucasagomes | NobodyCam, same error w/o oslo.messaging? | 17:25 |
| *** zdin0bot1 has joined #openstack-ironic | 17:26 | |
| NobodyCam | I have gotten that with both... I believe that is a real random error. just happen to be power syncing | 17:26 |
| *** vkozhukalov has quit IRC | 17:26 | |
| Shrews | woohoo! go, go shrews3 http://paste.openstack.org/show/77116/ | 17:27 |
| * Shrews crosses fingers and toes | 17:27 | |
| *** zdin0bot has quit IRC | 17:27 | |
| jbjohnso | devananda, let me look | 17:27 |
| *** russellb is now known as rustlebee | 17:28 | |
| jbjohnso | devananda, so ipmitool doesn't support getting the device? | 17:28 |
| NobodyCam | brb | 17:29 |
| jbjohnso | lucasagomes, oh, let me explain that comment | 17:30 |
| jbjohnso | lucasagomes, it's not by device that 'getting' is a problem | 17:30 |
| jbjohnso | historically, I've seen systems that don't implement getting *or* setting | 17:31 |
| jbjohnso | under the covers, it's impossible at the IPMI layer to support set and not get of that parameter | 17:31 |
| jbjohnso | well, guess it's not impossible, but man that would be some weird work to make it one way... | 17:31 |
| *** romcheg1 has joined #openstack-ironic | 17:32 | |
| jbjohnso | I have seen some that errored on any boot device request, and I've seen some that happliy except it, but the respective boot firmware does not honor it | 17:32 |
| *** matty_dubs|lunch is now known as matty_dubs | 17:32 | |
| jbjohnso | but I see now that ipmitool doesn't actually front end getting the current setting.. | 17:33 |
| lucasagomes | that's a good feeback | 17:33 |
| jbjohnso | raw 0 9 5 0 0 | 17:33 |
| jbjohnso | that would be the ipmitool way to get the data | 17:33 |
| jbjohnso | 01 05 80 00 00 00 00 | 17:33 |
| jbjohnso | the return code is implicitly 0 in that case... | 17:33 |
| lucasagomes | yeah, well, I gotta dig more into the ipmitool, but I didn't find the option the last time I tried | 17:33 |
| jbjohnso | and then you have to convert and bitmask the values | 17:34 |
| jbjohnso | or chassis bootparam get 5 | 17:35 |
| jbjohnso | - Boot Device Selector : Force PXE | 17:36 |
| lucasagomes | right will give it a try :D | 17:36 |
| JayF | jbjohnso: +1 I've seen similar weirdness with setting boot devices | 17:37 |
| jroll | so while we're talking ipmi stuff | 17:37 |
| jbjohnso | devananda, fwiw, I updated for ipmi 2.0 rev 1.1, though the most interesting bits (lan cfg params) aren't done at all | 17:37 |
| jroll | I have a question | 17:37 |
| devananda | jbjohnso: gotcha. so most likely, either the BMC supports both or neither, and we may or may not even know (eg, silent accept-and-ignore is possible) | 17:37 |
| jbjohnso | devananda, correct. The BMC is nothing more than 'remember these bits' | 17:38 |
| jbjohnso | devananda, it's really BIOS/UEFI that asks BMC for those bits, and if it don't bother, well then the obvious happens | 17:38 |
| jbjohnso | but that's a problem of using a less than thorough vendor, if it doesn't work in a silent fashion, tht's between a user and his vendor | 17:39 |
| jbjohnso | devananda, btw, how's your broadcom nic firmware going ;) | 17:39 |
| devananda | jbjohnso: right. so if getbootdev returns None -- that probably means the BMC just doesn't know. | 17:39 |
| *** kathay has quit IRC | 17:39 | |
| jbjohnso | devananda, it means 'no specific guidance' | 17:39 |
| jroll | we have a vendor that has a bug in set boot device - they recommend sending a sequence of raw bytes before sending the bootdev command. is that possible to upstream? maybe make it some kind of config option? | 17:39 |
| jbjohnso | devananda, I call that 'default' | 17:39 |
| jbjohnso | jrist, what is the sequence of raw bytes? | 17:39 |
| jbjohnso | jroll, err what is the sequence of raw bytes? | 17:40 |
| jroll | I don't remember off hand | 17:40 |
| jroll | sec | 17:40 |
| jrist | jbjohnso: :) | 17:40 |
| jbjohnso | now everyone knows exactly how many characters I type before I hit 'tab' | 17:40 |
| devananda | jroll: my guess is, that would fit well within pyghmi but not ironic | 17:40 |
| jrist | I was just going to spout off some random bytes | 17:40 |
| jroll | lol | 17:40 |
| jroll | devananda: hmm | 17:40 |
| jroll | we're using ipmitool atm | 17:40 |
| jbjohnso | right, but if I see the bytes, I might be able to understand what they are getting at | 17:40 |
| jroll | JoshNang couldn't get pyghmi working with our hardware iirc | 17:40 |
| jroll | oh sorry, deva distracted me | 17:41 |
| jroll | 00 08 03 08 | 17:41 |
| jroll | jbjohnso: ^ | 17:41 |
| JayF | that's the combination to my luggage! | 17:41 |
| jbjohnso | ok, one moment | 17:41 |
| comstud | I just used that joke last night. | 17:41 |
| jroll | heh | 17:41 |
| JoshNang | :) | 17:42 |
| devananda | jbjohnso: so one of my req's (i may have mentioned this to you a while back) to getting pyghmi as the default power driver in ironic | 17:42 |
| jbjohnso | I'm combing through the spec as we speak (to keep with the theme) | 17:42 |
| devananda | jbjohnso: is going to be automated testing on hardware | 17:42 |
| NobodyCam | ready kafka | 17:42 |
| jbjohnso | set system boot options.. | 17:42 |
| JoshNang | jroll: i didn't debug at all why pyghmi wasn't working on our hardware. i tried, it didn't immediately work (think it had problems initiating a connection), and ipmitool filled all our needs, so i left it there. | 17:43 |
| jroll | JoshNang: heh | 17:43 |
| jroll | ok | 17:43 |
| jbjohnso | jroll, ok, pyghmi already does taht | 17:43 |
| jroll | jbjohnso: sends those bytes with the bootdev command? | 17:43 |
| jbjohnso | jroll, yep | 17:43 |
| jroll | huh. | 17:43 |
| jroll | awesome | 17:44 |
| jbjohnso | jroll, it disables a very annoying feature of set boot device | 17:44 |
| jbjohnso | 'don’t automatically clear boot flag valid bit if Chassis Control command not | 17:44 |
| jbjohnso | received within 60-second timeout (countdown restarts when a Chassis | 17:44 |
| jbjohnso | Control command is received)' | 17:44 |
| jbjohnso | I hate that behavior and hate that it is supposed to be default | 17:44 |
| jbjohnso | so pyghmi says 'screw it' and already does that | 17:44 |
| jbjohnso | so I agree with your vendor | 17:44 |
| jbjohnso | well, it might be a bug still yet | 17:45 |
| jroll | heh | 17:45 |
| jroll | cool | 17:45 |
| jbjohnso | they may expire the timer near end of post instead of waiting for chassis reset | 17:45 |
| *** lucasagomes is now known as lucas-dinner | 17:45 | |
| jroll | it's definitely a bug, it messes with the sata controller somehow. but anyway. cool. | 17:46 |
| jroll | thanks :) | 17:46 |
| jbjohnso | anyhow, pyghmi better work or else someone (me) has failed. Unless it's a certain dell model in which case I'll hate them for a blatant security violation | 17:46 |
| jroll | lol | 17:46 |
| jroll | yeah, I'll poke at it today | 17:46 |
| jroll | and file bugs as needed | 17:46 |
| jbjohnso | can't remember which one it was, but they made man-in-the-middle prevention impossible in some firmware on some model | 17:46 |
| jroll | lovely | 17:46 |
| jroll | thanks dell :P | 17:46 |
| jbjohnso | I hope that day is long gone though | 17:47 |
| jbjohnso | If you do see a failure, and wouldn't mind capturing it and divulging the password used for the session captured | 17:47 |
| jbjohnso | with tcpdump or wireshark | 17:47 |
| jbjohnso | then I can decrypt it and tell roughly what might be weird | 17:48 |
| jroll | sure thing | 17:48 |
| jbjohnso | I tried even avoiding the stuff that demands 'intelplus' in ipmitool by being sneaky | 17:48 |
| *** Hefeweizen has quit IRC | 17:50 | |
| jbjohnso | jroll, generally, I can make sense of 'weird random byte streams' in the context of ipmi at least | 17:52 |
| jbjohnso | my goal is to have reasonable behavior out of the box without configuration if at all possible. I'm annoyed at lan v lanplus and intelplus and so on | 17:52 |
| jroll | jbjohnso: cool, I'm glad one of us can make sense of that :) | 17:53 |
| jbjohnso | *SO FAR* I've been able to avoid most of the drivers of those options through careful planning... | 17:53 |
| jbjohnso | well, 'careful' may be charitable.. | 17:53 |
| *** hemna has joined #openstack-ironic | 17:53 | |
| jroll | heh | 17:54 |
| jbjohnso | jroll, response = self.raw_command(netfn=0, command=8, data=(3, 8)) | 17:54 |
| jbjohnso | see, it was already in there... | 17:54 |
| jbjohnso | easiest feature request ever | 17:54 |
| jroll | I believed you :) | 17:54 |
| * jroll goes off to test | 17:55 | |
| jbjohnso | I'll claim that I considered the request so important, I actually time travelled (backwards) to fix it | 17:55 |
| jbjohnso | time travelling forwards is too easy... | 17:55 |
| jbjohnso | well, at the rate of 1 s/s at least | 17:56 |
| jroll | lol | 17:56 |
| *** hemna has quit IRC | 17:57 | |
| jbjohnso | devananda, fyi, I am able to pull all the sensor data with pyghmi at the moment | 17:58 |
| jbjohnso | devananda, and provided a 'health' function to get to the 'what is actually wrong' bit of the output (if anything) | 17:58 |
| jbjohnso | I still need to do the inventory and eventlog bits, but my next stop is socket pooling and maybe multiprocessing pooling... | 17:58 |
| devananda | jbjohnso: inventory and console/event log would be great | 17:59 |
| jbjohnso | devananda, yeah, the biggest thing about console is the security model you want | 18:00 |
| jbjohnso | devananda, since I assume you don't want to be in the middle of a highly stateful connection... | 18:00 |
| devananda | jbjohnso: nope. and fwiw, i think we need console *log* more than an interactive console | 18:01 |
| devananda | jbjohnso: but others may disagree | 18:01 |
| JayF | I disagree | 18:01 |
| devananda | see :) | 18:01 |
| NobodyCam | devananda: not I | 18:01 |
| jbjohnso | devananda, well, confluent offers both | 18:01 |
| JayF | An interactive console is super useful to recovering busted machines without needing administrator intervention | 18:01 |
| jbjohnso | it just requires some authentication scheme for the live console | 18:01 |
| JayF | so a user consuming a node from Ironic can fix their machine if, say, they boot a kernel on it that bricks it :) | 18:01 |
| jbjohnso | the plain text console is indeed plain text, but there is a binary metadata file that provides extra awesomeness | 18:01 |
| NobodyCam | log would be awesome for the gate tests where you can not do interav=ctive | 18:01 |
| NobodyCam | interactive even | 18:02 |
| JayF | interactive > log > nothing | 18:02 |
| jbjohnso | err the console log | 18:02 |
| devananda | JayF: well, if they deploy a bad kernel, they should just deploy another good kernel | 18:02 |
| jbjohnso | devananda, what happens if they flash that HP provided broadcom firmware ;) | 18:02 |
| JayF | devananda: I mean more in the sense of a user, in an operating system, doing something like "apt-get dist-upgrade", and the resulting kernel not booting | 18:02 |
| jbjohnso | heh | 18:02 |
| jbjohnso | anyway | 18:02 |
| devananda | jbjohnso: right... heh. | 18:02 |
| jbjohnso | https://www.youtube.com/watch?v=G_lDaktYnsQ | 18:03 |
| NobodyCam | lucas-dinner: with out patch :( http://paste.openstack.org/show/ZZh2uWojDvFsr9d9AEZN/ | 18:03 |
| jbjohnso | that's how confluent can do console | 18:03 |
| devananda | JayF: "nova rescue $INSTANCE" | 18:03 |
| jbjohnso | and there is logging in that | 18:03 |
| NobodyCam | retesting with the patch | 18:03 |
| devananda | JayF: to PXE boot the instance with a known-good kernel so they can mount / and fix their grub conf (or what ever) | 18:03 |
| devananda | JayF: no need for SOL | 18:03 |
| jroll | jbjohnso: another feature request: allow ctrl-c to interrupt ipmi operations :) | 18:03 |
| JayF | Hmm. That sounds awesome. I wonder if we support that in the Rackspace cloud... | 18:03 |
| JayF | my Openstack knowledge is shaded a bit more red and black than it should be | 18:04 |
| jroll | jbjohnso: for example if they hang | 18:04 |
| devananda | JayF: hah. i dunno, but it's part of nova's API and should be easy enough to implement in ironic | 18:04 |
| jbjohnso | jroll, ctrl-c doesn't interrup something? | 18:05 |
| jroll | jbjohnso: I have a get_power() that hung and ctrl-c doesn't help :/ | 18:05 |
| jroll | I'm in the python shell fwiw | 18:05 |
| jbjohnso | devananda, easiest thing is to give console. Even if there is a 'better' way, it's easier to do what is familiar | 18:05 |
| jbjohnso | jroll, hmm, but you were able to make the object ok? | 18:05 |
| NobodyCam | brb... quick walkies | 18:05 |
| jbjohnso | jroll, what about the ipmitcl.py util? | 18:05 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 18:06 |
| jroll | jbjohnso: yeah, I did: set_bootdev(), set_power(), get_power(), set_power(), get_power() #hung | 18:06 |
| jbjohnso | devananda, besides, there are also vendors that don't make some things available except via F1 | 18:06 |
| jroll | jbjohnso: I haven't tried that (nor was I aware of it) | 18:06 |
| jbjohnso | which is in no way a good situation, but a console can alleviate the bad... | 18:06 |
| jroll | I think this is just a random bmc oddity | 18:06 |
| jbjohnso | jroll, let me see what happens | 18:06 |
| jroll | we've had plenty of those even with ipmitool | 18:06 |
| jroll | jbjohnso: hm? | 18:07 |
| jbjohnso | oh, I'm doing the python interpreter | 18:07 |
| jbjohnso | to see if I can see the same thing | 18:07 |
| jroll | oh | 18:07 |
| jbjohnso | I actually never used it from the interpreter directly, believe it or not | 18:07 |
| jroll | I mean, I'm definitely inclined to blame our hw for the lockup, but ctrl-c wasn't working yeah | 18:07 |
| jroll | ha | 18:07 |
| jroll | heh, it did end up booting | 18:08 |
| jbjohnso | set_power to what? | 18:08 |
| jroll | seems to wfm | 18:08 |
| jroll | I flipped it off, then on | 18:09 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 18:09 |
| jroll | then get_power after on hung | 18:09 |
| jroll | but it did boot | 18:09 |
| jbjohnso | jroll, shared port configuration? | 18:09 |
| jbjohnso | (wondering if the on caused link state change to have an outage...) | 18:09 |
| jroll | the OOB NIC also has a 1g NIC, yeah | 18:10 |
| jbjohnso | though that should raise an exception | 18:10 |
| jroll | we *usually* don't see a hang using ipmitool | 18:10 |
| jroll | idk | 18:10 |
| * jroll kicks $vendor | 18:10 | |
| jbjohnso | wondering if it's specifically related to the timing of the power cycle and network conditions... | 18:11 |
| jbjohnso | a session that is valid but loses networking for a blip should still timeout or recover... | 18:11 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Don't log MACAlreadyExists in Manager.update_port https://review.openstack.org/90436 | 18:11 |
| jbjohnso | jroll, but it did seem to at least generally 'work as advertised'? | 18:11 |
| jroll | yeah | 18:11 |
| jroll | yep :) | 18:11 |
| JayF | Ah so the difference between CLI ipmitool and pyghmi in this case is that the session is persistent? | 18:11 |
| jroll | JayF: yeah | 18:11 |
| jroll | although how ironic uses pyghmi is one session per command | 18:12 |
| jbjohnso | JayF, yeah, the session hangs on for dear life for the lifetime of the object | 18:12 |
| jbjohnso | incidentally, it also caches SDR to memory per session | 18:12 |
| jroll | jbjohnso: fwiw it worked this time, so who knows | 18:12 |
| jbjohnso | well, SDR is fetched the first time on demand, but then retained | 18:12 |
| devananda | jroll: your OOB and IB NICs share a connection? | 18:13 |
| JayF | Well, for now. | 18:13 |
| jbjohnso | so the first 'get_health' is significantly longer than a follow up 'get_health' on the object | 18:13 |
| jroll | devananda: not really | 18:13 |
| jroll | devananda: they won't | 18:13 |
| devananda | 18:10:07 < jroll> the OOB NIC also has a 1g NIC, yeah | 18:13 |
| jroll | they do today | 18:13 |
| jroll | right | 18:13 |
| devananda | ah | 18:13 |
| jroll | we also have other NICs on the box | 18:13 |
| devananda | jroll: yay security | 18:13 |
| jroll | that 1g should be disabled | 18:13 |
| devananda | right | 18:13 |
| jroll | right now it is not | 18:13 |
| devananda | k | 18:13 |
| jroll | :) | 18:14 |
| jroll | so here's a thing | 18:14 |
| jroll | say a BMC gets in a state where it needs a reset | 18:14 |
| jbjohnso | jroll, was the command perhaps longer than 60 seconds at the prompt? | 18:14 |
| jbjohnso | jroll, in the interpreter, while at the prompt, our keepalive isn't running and I think the bmc got mad... | 18:15 |
| jroll | will ironic end up putting the node in maintenance mode or? | 18:15 |
| jroll | jbjohnso: ahhh :) | 18:15 |
| jbjohnso | jroll, I just went back and my object needed to be ctrl-ced a lot before it gave up | 18:15 |
| devananda | jroll: atm, i think it will, because sync_power_state will fail repeatedly | 18:15 |
| jbjohnso | ^C^C^C^C^C^C^CTr | 18:15 |
| jbjohnso | 'Tr' is the Traceback from KeyboardInterrupt | 18:15 |
| jroll | huh | 18:16 |
| jroll | devananda: ok | 18:16 |
| jbjohnso | >>> cmd.get_health() | 18:16 |
| jbjohnso | {'badreadings': [{'name': 'Mezz Exp 2 Fault', 'imprecision': None, 'value': None, 'states': ['Critical'], 'health': 2, 'units': '', 'type': 'Add-in Card'}], 'health': 2 | 18:16 |
| jroll | cool | 18:16 |
| jbjohnso | now if I wait long enough, I bet it gets hung up... | 18:16 |
| *** zdin0bot1 has quit IRC | 18:17 | |
| jbjohnso | I'll have to leave it at prompt for 60 seconds or so... | 18:17 |
| *** zdin0bot has joined #openstack-ironic | 18:17 | |
| jbjohnso | I wouldn't have noticed this because usually the event loop is running... | 18:17 |
| matty_dubs | linggao: About to +1 your patch. But, funny story... Was testing ipmitool sol activate command over ssh, vs. in shellinaboxd, to make sure it worked before testing your patch | 18:19 |
| matty_dubs | ssh also binds to ~?, and ~. disconnects you from an ssh session apparently | 18:19 |
| matty_dubs | Took me a little bit to figure out why I kept losing my ssh connection when testing. | 18:19 |
| linggao | Hi matty_dubs, yes. I encountered that problem before. :) | 18:20 |
| linggao | I'll add -v flag to the ipmitool commnad if debug is enabled per Yuriy's comment. So please wait... | 18:21 |
| jbjohnso | matty_dubs, that's why 'confetty' doesn't bind '~' for the escape sequences ;) | 18:21 |
| matty_dubs | linggao: Sure thing | 18:21 |
| matty_dubs | jbjohnso: Yay! ;) | 18:22 |
| jbjohnso | I love it when I play 'I forgot how many hops I sshed through, guess the number of ~s' | 18:22 |
| matty_dubs | It's been years since I remembered the ssh escape sequences, and I was a couple levels deep in ssh sessions, so I was completely baffled when they all just closed while testing this | 18:23 |
| jbjohnso | jroll, btw, ctrl-c once did work when I waiting for it to catch... I don't know | 18:23 |
| jbjohnso | well, I'll not pay too close attention, since I'm pretty sure that leaving the object idle at the interactive prompt isn't a significant use case.. | 18:23 |
| jbjohnso | otherwise I'd start doing things that could wreak havoc with eventlet/non-eventlet environments... | 18:24 |
| openstackgerrit | linggao proposed a change to openstack/ironic: Support serial console access https://review.openstack.org/64100 | 18:33 |
| matty_dubs | Is pep8 going to gripe about line 430? https://review.openstack.org/#/c/64100/24..25/ironic/drivers/modules/ipmitool.py | 18:35 |
| openstackgerrit | linggao proposed a change to openstack/ironic: Support serial console access https://review.openstack.org/64100 | 18:35 |
| matty_dubs | That was fast! ;) | 18:36 |
| linggao | matty_dubs, I forgot to run pep8. This new one should pass the Jekins. | 18:37 |
| lucas-dinner | NobodyCam, :( | 18:37 |
| NobodyCam | lucas-dinner: testing again with patch | 18:37 |
| linggao | matty_dubs, thanks a lot for testing. | 18:37 |
| matty_dubs | My pleasure | 18:37 |
| *** athomas has quit IRC | 18:38 | |
| adam_g | oh finally | 18:43 |
| adam_g | check-tempest-dsvm-virtual-ironic SUCCESS in 23m 30s (non-voting) | 18:43 |
| NobodyCam | adam_g: \o/ | 18:43 |
| devananda | adam_g: woot! | 18:44 |
| adam_g | should we make that voting at some point in the near future, if it proves stable over the next week or so? | 18:44 |
| NobodyCam | adam_g: ++ | 18:45 |
| *** comstud is now known as bearhands | 18:49 | |
| devananda | adam_g: yes | 18:49 |
| devananda | adam_g: we should make 2 changes: a) voting on ironic and python-ironic, b) move out of experimental for devstack, devstack-gate, and nova | 18:49 |
| adam_g | devananda, yeah.. im also going to add it to DIB. since a change there earlier this week broke our deployment ramdisk | 18:53 |
| *** zdiN0bot1 has joined #openstack-ironic | 19:03 | |
| *** vkozhukalov has joined #openstack-ironic | 19:06 | |
| *** zdiN0bot1 has quit IRC | 19:07 | |
| *** datajerk has quit IRC | 19:10 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 19:13 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 19:13 |
| devananda | adam_g: ++ | 19:15 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests DO NOT MERGE https://review.openstack.org/85529 | 19:16 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Workaround race with Neutron port updates https://review.openstack.org/84361 | 19:16 |
| devananda | NobodyCam: still testing https://review.openstack.org/#/c/88307/ ? | 19:16 |
| NobodyCam | yes :( | 19:16 |
| *** zdin0bot1 has joined #openstack-ironic | 19:17 | |
| *** zdin0bot has quit IRC | 19:17 | |
| NobodyCam | i forgot to save the images and overwrote them. :( so needed to rebuild | 19:17 |
| NobodyCam | which is not all that quick atm | 19:17 |
| NobodyCam | its runing right now fyi | 19:18 |
| *** datajerk has joined #openstack-ironic | 19:23 | |
| *** zdin0bot1 has quit IRC | 19:25 | |
| *** zdin0bot has joined #openstack-ironic | 19:25 | |
| adam_g | https://review.openstack.org/#/c/90452/ | 19:28 |
| adam_g | devananda, ^ | 19:28 |
| *** zdin0bot has quit IRC | 19:31 | |
| *** igor has quit IRC | 19:36 | |
| *** zdin0bot has joined #openstack-ironic | 19:45 | |
| *** jistr has quit IRC | 19:45 | |
| *** zdin0bot has quit IRC | 19:46 | |
| openstackgerrit | A change was merged to openstack/ironic: Spawn support for TaskManager and 2 locking fixes https://review.openstack.org/89328 | 19:46 |
| *** harlowja is now known as harlowja_away | 19:49 | |
| *** shakamunyi has quit IRC | 19:50 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Remove Heartbeat-Before check from heartbeat https://review.openstack.org/90457 | 19:57 |
| *** harlowja_away is now known as harlowja | 20:09 | |
| JayF | Passthru just got approved! \o/ | 20:13 |
| *** jdob has quit IRC | 20:15 | |
| *** jdob has joined #openstack-ironic | 20:15 | |
| *** ifarkas has quit IRC | 20:20 | |
| *** jdob has quit IRC | 20:24 | |
| *** jdob has joined #openstack-ironic | 20:24 | |
| openstackgerrit | A change was merged to openstack/ironic: Drivers may expose a top-level passthru API https://review.openstack.org/81919 | 20:28 |
| *** igor has joined #openstack-ironic | 20:30 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Remove Heartbeat-Before check from heartbeat https://review.openstack.org/90457 | 20:31 |
| jroll | GheRivero: around? | 20:34 |
| *** igor has quit IRC | 20:34 | |
| devananda | bearhands: what'd happen if a compute node is deleted in nova, while there are instance(s) on it? | 20:36 |
| bearhands | hm | 20:36 |
| bearhands | 'nothing' ? | 20:36 |
| bearhands | the instances would be like orphaned, I guess | 20:37 |
| bearhands | I think you can still delete them | 20:37 |
| devananda | but nova wont try to delete them automatically? | 20:37 |
| NobodyCam | he's utc +/- 1??? | 20:37 |
| devananda | that seems odd :) | 20:37 |
| bearhands | well | 20:37 |
| bearhands | i don't think there's anything that deletes the node in nova | 20:37 |
| bearhands | i think i saw recently someone was adding an admin API call for it | 20:38 |
| devananda | http://git.openstack.org/cgit/openstack/nova/tree/nova/compute/manager.py#n5459 | 20:38 |
| bearhands | but | 20:38 |
| GheRivero | jroll: pong | 20:38 |
| bearhands | yeah | 20:38 |
| bearhands | so that would be for baremetal only | 20:38 |
| GheRivero | devananda: congrats! | 20:38 |
| bearhands | (and now ironic, I guess) | 20:38 |
| devananda | bearhands: if eg. nova.virt.ironic.driver:get_available_nodes() returns an empty list suddenly, the rt would delete everything | 20:38 |
| devananda | yea | 20:38 |
| devananda | GheRivero: thanks! | 20:38 |
| bearhands | devananda: right, don't do that | 20:39 |
| devananda | heh | 20:39 |
| jroll | GheRivero: hey, wanted to let you know I just addressed your question here https://review.openstack.org/#/c/88722/ :) | 20:39 |
| bearhands | that's kind of like... | 20:39 |
| bearhands | you keep some state on your openstack instances. you write something that polls nova-api for your instances | 20:39 |
| bearhands | and destroy the data if you see the instances gone | 20:40 |
| jroll | GheRivero: sorry I missed that before | 20:40 |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Use systemd-nspawn instead of docker for image https://review.openstack.org/90465 | 20:40 |
| bearhands | and nova-api randomly returns [] ? | 20:40 |
| devananda | https://review.openstack.org/#/c/88590/ | 20:40 |
| devananda | NobodyCam: please take a moment to look at my comment ^ | 20:40 |
| bearhands | (we actually had this problem pre-openstack :) | 20:40 |
| GheRivero | jroll: thanks. will review it soon! | 20:40 |
| devananda | bearhands: that's pretty much what i thought. thanks :) | 20:40 |
| NobodyCam | oh | 20:41 |
| NobodyCam | revoving /me's vote | 20:41 |
| NobodyCam | removing even | 20:41 |
| bearhands | devananda: at least here, I don't think much badness happens... the node will get recreated | 20:41 |
| bearhands | when it shows up again | 20:42 |
| devananda | bearhands: and the instance that was on it and associated with the previous compute node's id? | 20:42 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Check configdrive size before writing to partition https://review.openstack.org/89390 | 20:43 |
| bearhands | devananda: instance would still be around.. instance is associated with host only by its name | 20:43 |
| bearhands | well | 20:43 |
| bearhands | instance.host, instance.node == strings | 20:43 |
| devananda | ah, there's no FK or reference from instance to compute_node ? | 20:44 |
| bearhands | not that i'm aware of | 20:44 |
| devananda | k | 20:44 |
| devananda | so anyway, it's not a good idea, but maybe not as bad as it looks | 20:44 |
| bearhands | right | 20:44 |
| bearhands | "don't do that"... but I think nova will handle it mostly okay. | 20:44 |
| bearhands | we could make nova a little smarter | 20:45 |
| bearhands | say | 20:45 |
| bearhands | don't delete the node unless it's been gone for 'x' minutes. | 20:45 |
| bearhands | or whatever | 20:45 |
| *** openstackstatus has quit IRC | 20:45 | |
| devananda | or delete the node if ironic returns valid data not including that node | 20:45 |
| devananda | (but dont do anyhting hasty when ironic doesn't return any data eg. due to network failure) | 20:46 |
| bearhands | or that, perhaps.. maybe an emptylist would be a warning that.. .maybe something else bad is going on | 20:46 |
| bearhands | :) | 20:46 |
| *** openstackstatus has joined #openstack-ironic | 20:46 | |
| bearhands | i don't see much reason to delete the entry immediately | 20:46 |
| devananda | raise an exception seems better to me. it'll be logged, and there shouldn't be any silent db changes as a result afaict | 20:47 |
| bearhands | scheduling will not use it after x seconds | 20:47 |
| bearhands | if it stops updating | 20:47 |
| bearhands | although | 20:47 |
| bearhands | i guess that may not be true with baremetal/ironic | 20:47 |
| bearhands | since the nova-compute service entry would be 'up' | 20:47 |
| devananda | right | 20:47 |
| bearhands | we could fix that | 20:47 |
| devananda | and >1 n-cpu will be reporting the same "host" name and the same set of nodes | 20:47 |
| bearhands | i think there's a hypervisor_mainteannce or something | 20:48 |
| * bearhands checks | 20:48 | |
| devananda | we're going to really need to sort the >1 n-cpu issues in atlanta | 20:48 |
| bearhands | hm, there's no way to label a ComputeNode entry as 'disabled'. | 20:48 |
| devananda | bearhands: adam_g has proposed to fake that by setting its resource counts to 0 | 20:49 |
| devananda | https://review.openstack.org/#/c/88476/4 | 20:49 |
| bearhands | yeah | 20:49 |
| *** zdiN0bot has joined #openstack-ironic | 20:49 | |
| bearhands | i thin i suggested the opposite... | 20:49 |
| devananda | o.0 | 20:49 |
| bearhands | faking its full | 20:49 |
| devananda | heh | 20:49 |
| bearhands | well | 20:49 |
| bearhands | kind of the same thing | 20:49 |
| bearhands | :) | 20:49 |
| devananda | iirc you said "dont delete it" | 20:50 |
| devananda | or someone did | 20:50 |
| bearhands | don't recall that | 20:50 |
| bearhands | well | 20:50 |
| *** openstackstatus has quit IRC | 20:50 | |
| bearhands | i might have said that, because that's true | 20:50 |
| devananda | but yea, setting it to have 0 available resources (whether by current,max=0 or current=max) seems the right way | 20:50 |
| devananda | and of those two, (current,max=0) seems better because it is not lying to resource trackers about total consumed resources | 20:51 |
| devananda | whereas setting current=max (eg, while the node is in maintenance mode) is going to skew any resource usage calculation | 20:51 |
| bearhands | 2014-04-17 13:06:24 <comstud> devananda: ya, that's what I was going to say.. show them to compute manager as fully utilized | 20:51 |
| bearhands | 2014-04-17 13:10:43 <comstud> which is why i think you still present it.. but fully utilized (as it would be with an instance) | 20:51 |
| bearhands | heh | 20:51 |
| devananda | because there's no tenant using it | 20:51 |
| *** openstackstatus has joined #openstack-ironic | 20:51 | |
| bearhands | yeah | 20:51 |
| bearhands | i agree | 20:51 |
| * bearhands watches the rain come down | 20:52 | |
| devananda | speaking of the weather ... it's awesomely sunny outside... i'm going for a walk shortly :) | 20:54 |
| *** jgrimm has quit IRC | 20:54 | |
| bearhands | devananda: sunny? where are you at? | 20:55 |
| devananda | Sunny Seattle! | 20:55 |
| devananda | lol | 20:55 |
| bearhands | lol i see | 20:55 |
| bearhands | east sfbay here.. and it just poured... for about 60 seconds | 20:55 |
| devananda | heh | 20:55 |
| bearhands | or, since i'm from the midwest, I would just call it normal 'rain' | 20:56 |
| bearhands | rain here to people seems to mean even a slight mist | 20:56 |
| *** ChanServ sets mode: +v openstackstatus | 20:56 | |
| *** romcheg1 has quit IRC | 20:57 | |
| devananda | copying the chatlog to that review for later reference | 20:57 |
| bearhands | cools | 20:57 |
| bearhands | i'm looking it over | 20:57 |
| *** linggao has quit IRC | 20:58 | |
| -openstackstatus- NOTICE: Gerrit will be unavailable for a few hours starting at 1600 UTC on Monday April 28th for an upgrade. See https://wiki.openstack.org/wiki/GerritUpgrade | 20:59 | |
| devananda | https://review.openstack.org/#/c/87396/ needs another +2 | 21:00 |
| devananda | NobodyCam: you had previously looked at it, have time for another look today? | 21:00 |
| * NobodyCam looking | 21:00 | |
| bearhands | devananda: i'm not sure that 'get_available_nodes' should filter out NOSTATE. that could make a node 'disappear' | 21:04 |
| *** matty_dubs is now known as matty_dubs|gone | 21:04 | |
| NobodyCam | devananda: have you tested this? | 21:04 |
| NobodyCam | :) | 21:04 |
| bearhands | as long as get_availble_resources checks NOSTATE (like you suggest), then we won't schedule to it. | 21:04 |
| bearhands | but anyway, left that feedback | 21:04 |
| devananda | bearhands: NOSTATE indicates a node that was just registered with ironic whose power state has not been fetched from the BMC yet | 21:05 |
| devananda | bearhands: that'll either convert to a real state (ON, OFF) or ERROR after some retries | 21:05 |
| bearhands | ya, but say.. something/someone screws up and sets it back to NOSTATE? | 21:05 |
| devananda | hmm | 21:05 |
| bearhands | i don't see harm in it still being presented to nova | 21:05 |
| bearhands | but with 0 resources | 21:06 |
| devananda | i'd say, make it unschedulable -- but dont delete the compute record | 21:06 |
| devananda | right | 21:06 |
| devananda | thought that's what i was suggesting | 21:06 |
| bearhands | right | 21:06 |
| bearhands | well you did, in get_available_resources() | 21:06 |
| devananda | wait. is this deleting compute records for things in ERROR state? | 21:06 |
| bearhands | but it's also filtered out in get_available_nodes() right now | 21:06 |
| * devananda rereads | 21:06 | |
| bearhands | in this patch | 21:06 |
| bearhands | i think get_available_nodes() should just reutrn the full node_list we get from ironic | 21:06 |
| bearhands | leave the unschedule magic completely to get_available_resources() | 21:07 |
| devananda | ooh right | 21:07 |
| *** jbjohnso has quit IRC | 21:07 | |
| bearhands | https://review.openstack.org/#/c/88476/4/ironic/nova/virt/ironic/driver.py,unified line 291 | 21:07 |
| devananda | yea, i would remove all the state check in get_available_nodes | 21:07 |
| bearhands | cools, left it as feedback, just wanted to make sure i wasn't on crack | 21:08 |
| devananda | just return all nodes. let the scheduler see them, even if they have 0 resources | 21:08 |
| devananda | side note | 21:08 |
| devananda | we could pass the node.maintenance property (and others) up as compute node properties, then add a FilterScheduler to match on it | 21:09 |
| devananda | exclude things that way, instead of using the resource counts | 21:09 |
| bearhands | i was looking at that | 21:09 |
| devananda | there was a session proposal to add that | 21:09 |
| bearhands | i commented on it yesterady or so | 21:09 |
| devananda | for hardware model # or something | 21:09 |
| bearhands | 127 # extra_resources is a json string containing arbitrary | 21:09 |
| bearhands | 128 # data about additional resources. | 21:09 |
| bearhands | 129 extra_resources = Column(Text) | 21:09 |
| bearhands | was thinking that could be a place to put it | 21:10 |
| NobodyCam | devananda: atm its looking like I'll be +2'ing 88307 in about 5 min | 21:10 |
| bearhands | it's not really a 'resource', but.. there's not a real good other place to put it | 21:10 |
| bearhands | maybe 'stats' IS the right place (which is where resources go today) | 21:10 |
| devananda | NobodyCam: awesome | 21:10 |
| boris-42 | bearhands btw for storing json | 21:10 |
| boris-42 | bearhands there is a interesting way in sqlalchemy | 21:10 |
| devananda | bearhands: stats ++ | 21:10 |
| boris-42 | bearhands something like https://github.com/stackforge/rally/blob/master/rally/db/sqlalchemy/types.py#L89 | 21:11 |
| bearhands | I'd almost prefer just adding a 'maintenance' column to this table | 21:11 |
| bearhands | :) | 21:11 |
| bearhands | but anyway | 21:11 |
| boris-42 | bearhands so sqalclhemy will automatically do load() and dump() | 21:11 |
| boris-42 | bearhands and track status dict() was changed | 21:11 |
| boris-42 | bearhands to update column | 21:11 |
| bearhands | boris-42: Yeah, I saw this somewhere else, also | 21:12 |
| boris-42 | bearhands I think docs of sqlalchemy lol=) | 21:12 |
| bearhands | objects covers up most of the need for this | 21:12 |
| boris-42 | bearhands heh I am afraid a bit about performance of such approach | 21:12 |
| bearhands | since they should be your interface to dbapi, they loads/dumps and deal with stuff | 21:12 |
| boris-42 | bearhands I mean objects | 21:13 |
| bearhands | which approach? | 21:13 |
| bearhands | well | 21:13 |
| bearhands | I'm afraid of using more things in sqlalchemy.. for performance | 21:13 |
| bearhands | :) | 21:13 |
| boris-42 | bearhands but it's next ORM | 21:13 |
| bearhands | The json conversion has to happen somewhere | 21:13 |
| boris-42 | bearhands ORM on ORM | 21:13 |
| bearhands | I'd prefer it to happen in objects, I think | 21:13 |
| bearhands | somewhat, but sqlalchemy sucks, man | 21:13 |
| boris-42 | bearhands but we still use it =) | 21:14 |
| boris-42 | bearhands and I don't think that sqla sucks=) | 21:14 |
| bearhands | Yeah, but I have an alternative to it... mysqldb direct. | 21:14 |
| bearhands | which won't have this | 21:14 |
| boris-42 | bearhands it will | 21:14 |
| bearhands | unless i implement it within my mysqldb impl | 21:14 |
| boris-42 | bearhands cause OpenStack uses SQL in wrong way | 21:14 |
| bearhands | well, i won't argue that | 21:14 |
| boris-42 | bearhands so but these are different things | 21:15 |
| bearhands | but even using it the correct way, sqlalchemy performance sucks way worse than using mysqldb directly. | 21:15 |
| boris-42 | bearhands okay with this agree | 21:15 |
| bearhands | i have the numbers stashed away somewhere to prove it | 21:15 |
| boris-42 | bearhands I remember | 21:15 |
| boris-42 | =) | 21:15 |
| bearhands | nod | 21:15 |
| bearhands | therefore, I'd prefer to use less from sqlalchemy, not more | 21:15 |
| bearhands | :) | 21:15 |
| boris-42 | bearhands so someday you are guys going to get rid of using it hardcoding direct usage of it inside objects? | 21:16 |
| bearhands | (and yeah, i know, objects is basically a mini-orm on top of an orm in our case :) | 21:16 |
| bearhands | boris-42: I'm not sure I would go that far... We need to support more than mysql | 21:16 |
| bearhands | But I could see that 'mysql' ends up meaning to use 'mysqldb' instead of sql-a | 21:17 |
| boris-42 | bearhands so the mini-orm on orm will be worse perfromance=) | 21:17 |
| bearhands | right, but we don't have to use the ORM | 21:17 |
| bearhands | we could use core | 21:17 |
| bearhands | and it's really not any less efficient, actually, the way we're doing things right now | 21:18 |
| bearhands | i mean, tell me how objects doing json vs sql-a doing json is different | 21:18 |
| boris-42 | bearhands same same=) | 21:19 |
| bearhands | hehe | 21:19 |
| boris-42 | bearhands https://review.openstack.org/#/c/43151/ | 21:19 |
| boris-42 | bearhands yep that could be hidden inside objects | 21:19 |
| boris-42 | bearhands but I don't see big difference to make it in db.api | 21:19 |
| bearhands | yeah i saw this patch a long time ago | 21:19 |
| bearhands | we've now eliminated 'stats' being in a separate table, which solves a lot of the problems the original query had | 21:20 |
| boris-42 | bearhands yep that makes totally sense | 21:20 |
| boris-42 | bearhands btw there is a lot of issue | 21:20 |
| boris-42 | bearhands with joins | 21:20 |
| bearhands | but ya, this points out where sql-a is slower than it should be | 21:21 |
| boris-42 | bearhands like this one http://pavlovic.me/rally/glance_list.html | 21:21 |
| bearhands | glance joins all of the properties | 21:21 |
| boris-42 | bearhands in cinder & nova the same | 21:21 |
| bearhands | which are key/value table | 21:21 |
| boris-42 | bearhands not key/value | 21:21 |
| bearhands | i mean, i assume it joins | 21:21 |
| boris-42 | bearhands key/value/created_at/deleted_at/updated_at | 21:21 |
| bearhands | well, image_properties is technically name/value rows | 21:21 |
| bearhands | right | 21:21 |
| devananda | ok, really going for a walk before the sun hides again | 21:21 |
| devananda | bbiab | 21:21 |
| bearhands | devananda: cools, i'm out soon also | 21:21 |
| bearhands | enjoy | 21:21 |
| boris-42 | bearhands soo | 21:22 |
| bearhands | boris-42: yeah, it's extra bad | 21:22 |
| boris-42 | bearhands yep dates are terrible | 21:22 |
| boris-42 | bearhands in joins) | 21:22 |
| boris-42 | bearhands so nova/cinder the same story | 21:22 |
| boris-42 | bearhands with volumes and instances | 21:22 |
| *** zdiN0bot has quit IRC | 21:23 | |
| NobodyCam | export DIB_REPOLOCATION_diskimage_builder=https://review.openstack.org/openstack/diskimage-builder | 21:23 |
| NobodyCam | gah ww | 21:23 |
| bearhands | gotta run, laters all.. have a good weekend | 21:26 |
| NobodyCam | have a good weekend bearhands | 21:26 |
| *** jdob has quit IRC | 21:29 | |
| openstackgerrit | Jay Faulkner proposed a change to openstack/ironic-python-agent: Use systemd-nspawn instead of docker for image https://review.openstack.org/90465 | 21:29 |
| *** igor has joined #openstack-ironic | 21:30 | |
| openstackgerrit | Adam Gandelman proposed a change to openstack/ironic: Decouple state inspection and availability check https://review.openstack.org/88476 | 21:31 |
| *** vkozhukalov has quit IRC | 21:32 | |
| *** igor has quit IRC | 21:35 | |
| lucas-dinner | devananda, NobodyCam thanks for the reviews :D | 21:41 |
| lucas-dinner | I saw that the oslo.messaging was approved as well and failed to merge I will fix it | 21:42 |
| NobodyCam | lucas-dinner: what are you doing up :p lol | 21:43 |
| lucas-dinner | NobodyCam, lol I was watching some series | 21:43 |
| lucas-dinner | was a bit tired to go out | 21:44 |
| NobodyCam | :) | 21:44 |
| NobodyCam | I'll reapprove | 21:44 |
| lucas-dinner | :) cheers, just fixing the conflicts | 21:46 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Port to oslo.messaging https://review.openstack.org/88307 | 21:47 |
| *** coolsvap|afk has quit IRC | 21:47 | |
| NobodyCam | :) | 21:48 |
| *** eguz has joined #openstack-ironic | 21:55 | |
| NobodyCam | stepping away for a bit will re +a after mr J is done | 21:57 |
| *** eghobo has quit IRC | 21:59 | |
| *** zdiN0bot has joined #openstack-ironic | 22:00 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 22:06 |
| openstackgerrit | A change was merged to openstack/ironic: nova.virt.ironic passes ephemeral_gb to ironic https://review.openstack.org/87396 | 22:23 |
| devananda | bearhands: if you're still around, 84823 needs a rebase. looks like it fell off my (and everyone's) queue | 22:23 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 22:24 |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Use systemd-nspawn instead of docker for image https://review.openstack.org/90465 | 22:25 |
| *** igor_ has joined #openstack-ironic | 22:30 | |
| devananda | Shrews: 88403 also needs a rebase, but yea, that should totally land. I didn't see it before ... | 22:33 |
| *** igor_ has quit IRC | 22:34 | |
| devananda | SpamapS: on the swap partition patch, any thoughts on whether that'll impact tripleo? | 22:38 |
| SpamapS | devananda: swap shmap | 22:41 |
| devananda | SpamapS: heh :) well, it changer partition order | 22:41 |
| SpamapS | devananda: We might be working around bugs in block device mappings that might cause us pain. Not sure. | 22:41 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Sync oslo logging https://review.openstack.org/87346 | 22:44 |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Remove 'fake' driver from default enabled drivers https://review.openstack.org/88711 | 22:44 |
| *** zdiN0bot has quit IRC | 22:46 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Factoring out PXE and TFTP functions https://review.openstack.org/90233 | 22:59 |
| *** zdiN0bot has joined #openstack-ironic | 23:00 | |
| *** tatyana has joined #openstack-ironic | 23:05 | |
| *** eguz has quit IRC | 23:22 | |
| *** eghobo has joined #openstack-ironic | 23:23 | |
| *** derekh has joined #openstack-ironic | 23:24 | |
| *** dkehn_ is now known as dkehnx | 23:30 | |
| *** igor has joined #openstack-ironic | 23:30 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 23:32 |
| *** igor has quit IRC | 23:35 | |
| *** derekh has quit IRC | 23:52 | |
| *** shakamunyi has joined #openstack-ironic | 23:55 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/89234 | 23:58 |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic-python-agent: Updated from global requirements https://review.openstack.org/88722 | 23:58 |
| *** mdenny has quit IRC | 23:58 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!