| *** openstack has joined #openstack-ironic | 13:49 | |
| jroll | morning ironic :) | 13:50 |
|---|---|---|
| agordeev | jroll: morning :) | 13:50 |
| *** pcrews has joined #openstack-ironic | 13:55 | |
| NobodyCam | good morning Ironic | 13:55 |
| jroll | do we have a bug/spec for being able to call delete on a build in progress? | 13:56 |
| jroll | or should that just work | 13:56 |
| jroll | hiya NobodyCam | 13:56 |
| NobodyCam | jroll: its never worked for me | 13:57 |
| jroll | I swear we had that written down | 13:57 |
| * jroll keeps looking | 13:58 | |
| *** geekyogi has joined #openstack-ironic | 13:58 | |
| jroll | a ha | 13:58 |
| jroll | https://blueprints.launchpad.net/ironic/+spec/make-tasks-interruptible | 13:58 |
| jroll | so this is a bit of a bug, too... if you delete an instance while still building, it seems to end up with provision_state 'active' | 13:59 |
| jroll | maybe | 14:00 |
| jroll | comstud found this last night, going to dig more today | 14:00 |
| NobodyCam | :) | 14:01 |
| *** bvivek has joined #openstack-ironic | 14:04 | |
| *** Poornima has quit IRC | 14:08 | |
| *** jgrimm has joined #openstack-ironic | 14:15 | |
| *** rwsu has joined #openstack-ironic | 14:16 | |
| NobodyCam | Haomeng: you here? | 14:25 |
| romcheg | Good morning NobodyCam! | 14:25 |
| NobodyCam | morning romcheg | 14:26 |
| NobodyCam | how was your weekend? | 14:26 |
| matty_dubs | Morning romcheg NobodyCam et al.! | 14:26 |
| romcheg | Morning matty_dubs! | 14:26 |
| NobodyCam | morning matty_dubs | 14:26 |
| *** geekyogi has quit IRC | 14:27 | |
| romcheg | NobodyCam: It's pretty good thanks! I'm hosting refugees from East here so it's quite interesting :) | 14:27 |
| *** geekyogi1 has joined #openstack-ironic | 14:27 | |
| NobodyCam | oh wow | 14:27 |
| romcheg | It's exam time for those who finish secondary schools | 14:27 |
| romcheg | s/finish/graduate | 14:28 |
| NobodyCam | :) | 14:28 |
| romcheg | They cannot pass exams on the East so they do that in Kharkiv | 14:28 |
| jroll | boris-42: around? | 14:29 |
| NobodyCam | romcheg: is this because of the stuff going on inthe east? | 14:29 |
| romcheg | NobodyCam: Exactly. | 14:29 |
| NobodyCam | :( | 14:29 |
| romcheg | NobodyCam: I have 3 students here in my apartment. Thanks God, Buddah, and all other possible gods that I don't have to pass those f**g exams anymore | 14:30 |
| NobodyCam | lol Oh so true | 14:31 |
| romcheg | NobodyCam: Now I look how worried they are and remember those External Independent Testing | 14:31 |
| *** Faizan has joined #openstack-ironic | 14:31 | |
| NobodyCam | yea. | 14:32 |
| romcheg | NobodyCam: It was hell. But now I don't give a s**t, they do :-З | 14:32 |
| romcheg | s/:-З/:-P | 14:33 |
| NobodyCam | hheehhe ya | 14:33 |
| NobodyCam | i have a question on: https://review.openstack.org/#/c/102435/24/specs/juno/send-data-to-ceilometer.rst lines:126 - 130 | 14:33 |
| NobodyCam | do we need specs for other drivers once this one is approves? | 14:34 |
| NobodyCam | approved even | 14:34 |
| NobodyCam | could not the other driver just ref this spec in their BP's | 14:35 |
| romcheg | NobodyCam: I don't feel like we do. This is an interface method which should behave the same in all drivers | 14:35 |
| NobodyCam | yea. i understand why he may not want to add it to the other driver like ipmiNative because they are unstable | 14:37 |
| NobodyCam | but the stubs should be added? | 14:37 |
| romcheg | NobodyCam: I'd say yes. If a method is in the interface the rest of the code assumes that all drivers implement it. So we have to either change the logic of the manager or add stubs to drivers | 14:39 |
| *** bvivek has quit IRC | 14:53 | |
| *** mitz- has quit IRC | 15:07 | |
| *** mitz has joined #openstack-ironic | 15:07 | |
| *** martyntaylor has quit IRC | 15:09 | |
| Faizan | Devananda: NobodyCam: thanks very much for your review comments on design spec for uefi support for irionic - https://review.openstack.org/#/c/99850 | 15:10 |
| *** mdorman has joined #openstack-ironic | 15:14 | |
| *** mitz has quit IRC | 15:15 | |
| *** ifarkas_ has quit IRC | 15:16 | |
| NobodyCam | :) | 15:16 |
| *** mitz has joined #openstack-ironic | 15:16 | |
| *** jcoufal has quit IRC | 15:18 | |
| *** lazy_prince is now known as killer_prince | 15:20 | |
| *** martyntaylor has joined #openstack-ironic | 15:25 | |
| *** Faizan has quit IRC | 15:30 | |
| *** rakesh_hs has quit IRC | 15:32 | |
| *** foexle has quit IRC | 15:33 | |
| *** dwalleck has joined #openstack-ironic | 15:35 | |
| *** athomas has quit IRC | 15:48 | |
| *** dwalleck_ has joined #openstack-ironic | 15:52 | |
| *** dwalleck_ has quit IRC | 15:53 | |
| *** max_lobur has joined #openstack-ironic | 15:53 | |
| NobodyCam | brb | 15:53 |
| devananda | g'morning, all | 15:53 |
| romcheg | Morning devananda! | 15:54 |
| *** dwalleck has quit IRC | 15:55 | |
| NobodyCam | good morning devananda | 15:56 |
| *** athomas has joined #openstack-ironic | 15:57 | |
| *** blamar has joined #openstack-ironic | 16:00 | |
| lucasagomes | morning NobodyCam devananda | 16:00 |
| NobodyCam | morning lucasagomes | 16:03 |
| NobodyCam | lucasagomes: I had a question on the ipmi spec | 16:05 |
| NobodyCam | - ipxe_enabled: Whether iPXE is enabled or not. | 16:05 |
| lucasagomes | NobodyCam, oh lemme take a look I have to update that spec | 16:05 |
| NobodyCam | will that be for the entire conductor | 16:05 |
| lucasagomes | NobodyCam, yes for the entire conductor | 16:06 |
| lucasagomes | NobodyCam, concerns? | 16:06 |
| NobodyCam | thinking... | 16:06 |
| lucasagomes | cause we are chainloading the ipxe boot file | 16:06 |
| lucasagomes | so even if it's a request from a dummy firmware ipxe will work | 16:07 |
| NobodyCam | was wounder / trying to find a edge case where we would need both pxe and ipxe | 16:07 |
| NobodyCam | but I dont think there is | 16:07 |
| lucasagomes | yeah, I think that would be messy as well | 16:07 |
| NobodyCam | ya | 16:08 |
| lucasagomes | another thing I wanna do next, after that spec would be to have a way to disable tftp completely | 16:08 |
| lucasagomes | if the machines can boot ipxe without chainloading | 16:08 |
| lucasagomes | no reason to have a tftp server | 16:08 |
| lucasagomes | everything can be done via http | 16:08 |
| NobodyCam | that would have to be per node | 16:08 |
| NobodyCam | as some hardware dose not support ipxe as defau;lt | 16:09 |
| lucasagomes | and maybe in the future maybe even the http server can go away, if we start putting things in glance swift :) | 16:09 |
| lucasagomes | imagine even the configs | 16:09 |
| lucasagomes | that would be neat | 16:09 |
| NobodyCam | ya | 16:10 |
| NobodyCam | lucasagomes: that should be our goal | 16:10 |
| lucasagomes | NobodyCam, yeah, but baby steps we need to start small | 16:10 |
| lucasagomes | and then we can build more stuff on top of it | 16:10 |
| NobodyCam | yep | 16:10 |
| boris-42 | jroll yep | 16:12 |
| boris-42 | jroll here | 16:12 |
| *** rameshg87 has joined #openstack-ironic | 16:15 | |
| *** matty_dubs is now known as matty_dubs|lunch | 16:16 | |
| rameshg87 | hello JoshNang | 16:18 |
| ndipanov | devananda, devananda, fwiw - destroy was already broken but was never getting hit since it would only happen on init_host which you override in ironic | 16:29 |
| ndipanov | devananda, you might want to audit the nova tests (not that it matters now) | 16:29 |
| devananda | ndipanov: wouldn't we hit destroy() simply during "nova delete" ? | 16:31 |
| *** ccrouch has joined #openstack-ironic | 16:31 | |
| ndipanov | not with the missing kw arg passed on | 16:31 |
| ndipanov | devananda, which is when it would break | 16:31 |
| devananda | ah | 16:31 |
| *** martyntaylor has left #openstack-ironic | 16:34 | |
| jroll | boris-42: hey, on osprofiler, any plans to do simpler metrics like counters and whatnot? | 16:35 |
| boris-42 | ? | 16:36 |
| boris-42 | jroll counters of what? | 16:36 |
| openstackgerrit | Nikola Dipanov proposed a change to openstack/ironic: Update destroy() driver method arguments https://review.openstack.org/103582 | 16:37 |
| jroll | boris-42: eg I want to know "this function was called x times" | 16:37 |
| boris-42 | jroll it can be done with simple aggregation | 16:38 |
| jroll | thinking something like graphite/statsd | 16:38 |
| jroll | how is the data stored? | 16:38 |
| boris-42 | jroll it's depends on many things | 16:38 |
| boris-42 | jroll in case of ceilometer it's bunch of notifications | 16:38 |
| boris-42 | jroll they can be transformed into tree | 16:38 |
| jroll | and is it pluggable if I don't want ceilometer? :) | 16:38 |
| Shrews | ndipanov: https://review.openstack.org/103137 is covering that missing param, too | 16:38 |
| Shrews | lucasagomes: ^^^^ | 16:39 |
| boris-42 | jroll I will try to make it not so hardocded | 16:39 |
| boris-42 | jroll osprofiler is plugable | 16:39 |
| boris-42 | jroll but integrating it inside openstack to make it work out of box is quite hard task=) | 16:39 |
| jroll | boris-42: ok, I need to look at this more | 16:39 |
| jroll | +1 | 16:39 |
| ndipanov | Shrews, bah OK will rebase on top of that | 16:39 |
| boris-42 | jroll so but as markmc dislike my patches | 16:40 |
| boris-42 | jroll in oslo.messaging I am going to make it a bit more plugable | 16:40 |
| * ndipanov silly docstrigs everywhere - ppl should write a novel if they like writing so much :) | 16:40 | |
| jroll | boris-42: ok, cool. will be watching :) | 16:40 |
| boris-42 | jroll but as I said it's important to get it with Ceilometer in | 16:41 |
| boris-42 | jroll so we will be able to use it in gates | 16:41 |
| boris-42 | jroll this is btw parsed profiling data https://gist.github.com/boris-42/c3c3ee1c2c7db40de236 | 16:41 |
| boris-42 | jroll for one trace | 16:41 |
| jroll | boris-42: sure, I agree there. I just prefer not to use ceilometer :) | 16:42 |
| jroll | thanks, on mobile but will check later | 16:42 |
| boris-42 | jroll so not sure that agree with this | 16:42 |
| boris-42 | jroll I am going to work on backend for profiler | 16:42 |
| * lucasagomes reading | 16:42 | |
| boris-42 | jroll something quite simple that will scale | 16:42 |
| jroll | boris-42: +1 for scalable sw | 16:43 |
| boris-42 | jroll e.g. just one table, that contains data that it need and it is well indexed | 16:43 |
| boris-42 | jroll this will scale very very | 16:43 |
| boris-42 | jroll even in sql | 16:43 |
| Shrews | adam_g`: I've tried pinging in -qa twice now for cores to look at my tempest change to no avail. Is there any one person in particular you ping in situations like this? | 16:43 |
| openstackgerrit | Nikola Dipanov proposed a change to openstack/ironic: Update destroy() driver method arguments https://review.openstack.org/103582 | 16:47 |
| lucasagomes | ndipanov, devananda , as Shrews pointed out https://review.openstack.org/103137 is adding that parameter + docstrings | 16:47 |
| ndipanov | lucasagomes, just updated it | 16:48 |
| ndipanov | lucasagomes, fwiw - tests pass for me locally (actually maybe not on top of your change lemme check) | 16:48 |
| lucasagomes | ndipanov, oh that was quick :) | 16:48 |
| *** chuckC has joined #openstack-ironic | 16:49 | |
| *** chuckC has left #openstack-ironic | 16:49 | |
| *** chuckC has joined #openstack-ironic | 16:49 | |
| *** ramineni has quit IRC | 16:50 | |
| openstackgerrit | Nikola Dipanov proposed a change to openstack/ironic: Add migrate_data keyword arg to driver destroy() https://review.openstack.org/103582 | 16:51 |
| *** coolsvap|afk is now known as coolsvap | 16:52 | |
| devananda | Shrews: mtreinish has been helpful for tempest/qa things, and seems not quite as busy as sdague | 16:52 |
| Shrews | devananda: he's MIA in -qa | 16:53 |
| ndipanov | ntw thanks lucasagomes Shrews | 16:53 |
| ndipanov | btw | 16:53 |
| JoshNang | rameshg87: hello! | 16:53 |
| rameshg87 | hello JoshNang | 16:53 |
| lucasagomes | yvw | 16:53 |
| rameshg87 | need some time of yours when you are free to discuss regarding swift temp url | 16:54 |
| JoshNang | sure! | 16:54 |
| rameshg87 | JoshNang, i had left a comment. did you have a chance to see that ? | 16:54 |
| rameshg87 | JoshNang, we would like to use the swift temp url mechanism for our swift objects as well (the objects not put by glance) | 16:55 |
| JoshNang | rameshg87: ah yes i saw that. | 16:55 |
| NobodyCam | brb | 16:56 |
| rameshg87 | JoshNang, hence i think we can move it to common/swift.py, and let image_service.py use the mechanism in swift.py | 16:56 |
| JoshNang | so, you should be able to generate swift temp urls using swiftclient. all the actual hashing and what not is going to be factored out to there (haven't removed it from my ironic patch yet) | 16:56 |
| rameshg87 | JoshNang, yeah the mechanism of generation of tempurl can be in swift.py and image_service.py can just call them | 16:57 |
| *** SpamapS_ is now known as SpamapS | 16:57 | |
| *** SpamapS has quit IRC | 16:57 | |
| *** SpamapS has joined #openstack-ironic | 16:57 | |
| JoshNang | rameshg87: thinking about it more, i'm not sure what code would go in swift.py other than a call to swiftclient | 16:57 |
| JoshNang | rameshg87: unless we want to add the url override features i have in the current glance patch. | 16:58 |
| JoshNang | (where, if for example, something doesn't have dns, you can override the swift cluster hostname with an ip address for your temp urls) | 16:58 |
| rameshg87 | JoshNang, i agree. but aren't all those related to swift more than glance. they apply for ironic-uploaded swift objects as well, right ? | 16:59 |
| JoshNang | rameshg87: so, the original intent and the easiest path is to use glance's direct_url to generate the tempurls | 16:59 |
| JoshNang | and i think that should be good for most people trying to generate a temp url for a glance object | 17:00 |
| *** matty_dubs|lunch is now known as matty_dubs | 17:00 | |
| rameshg87 | JoshNang, do you mean we can keep the current routines in image_service.py itself ? | 17:00 |
| JoshNang | rameshg87: well, the parts that deal with deciding what url to use for a given image_id, yes. for the "here's a url path, generate a temp url", that will live here: https://review.openstack.org/#/c/102632/ | 17:01 |
| devananda | lucasagomes: w.r.t. oslo.messaging - is there mroe work we need to do? | 17:02 |
| rameshg87 | JoshNang, i agree. | 17:02 |
| devananda | lucasagomes: it looks like all the version=X stuff landed, so maybe that's all done? | 17:02 |
| JoshNang | rameshg87: actually, one thing that probably needs to be shared is the temp url key | 17:02 |
| rameshg87 | JoshNang, the temp url key needn't be shared | 17:02 |
| lucasagomes | devananda, it's done | 17:02 |
| devananda | lucasagomes: great | 17:02 |
| *** bvivek has joined #openstack-ironic | 17:03 | |
| rameshg87 | JoshNang, glance account can have it's own key, the admin account can have another key | 17:03 |
| JoshNang | rameshg87: ah, that makes sense | 17:03 |
| *** derekh_ has quit IRC | 17:05 | |
| *** geekyogi1 has quit IRC | 17:06 | |
| rameshg87 | JoshNang, the review https://review.openstack.org/#/c/102632/1, i guess you have proposed a change in swiftclient for generating tempurls ? | 17:06 |
| JoshNang | rameshg87: yup! | 17:06 |
| lucasagomes | devananda, we would need to keep an eye on the reviews to see if those versions are being updated correctly, I looked a bit into make it checked programatically | 17:07 |
| lucasagomes | but I don't think that would be very easy/possible | 17:07 |
| rameshg87 | JoshNang, i do have a /usr/local/bin/swift-temp-url in my system | 17:07 |
| JoshNang | rameshg87: during the blueprint process, we decided "why wait to land it then factor it out"? | 17:07 |
| rameshg87 | JoshNang, how is this different from swift-temp-url ? | 17:07 |
| JoshNang | rameshg87: i believe that is install when you install swift, not swiftclient (at least when i check a couple months ago) | 17:07 |
| devananda | lucasagomes: right, but I think it's reasonable to expect all core reviewers to undersatnd hwo to do this properly | 17:09 |
| rameshg87 | JoshNang, so is your change not callable from the swift api mentioned here ? : http://docs.openstack.org/developer/python-swiftclient/swiftclient.html#module-swiftclient.client | 17:09 |
| JoshNang | rameshg87: it'd be something to the effect of swiftclient.tempurl.generate_temp_url(url, key, duration) | 17:10 |
| rameshg87 | JoshNang, i guess swiftclient.utils.generate_temp_url(url, key, duration) | 17:11 |
| JoshNang | rameshg87: err swiftclient.utils.generate_temp_url. i may add it to that client as well | 17:11 |
| rameshg87 | JoshNang, am i correct ? | 17:11 |
| JoshNang | rameshg87: yup! | 17:11 |
| rameshg87 | JoshNang, okay makes sense. so given a url, the functionality of generating tmp url can be moved to swiftclient | 17:11 |
| JoshNang | rameshg87: correct. | 17:11 |
| rameshg87 | JoshNang, so even the image_service.py can make use of this, right ? are you planning to change that ? | 17:12 |
| JoshNang | rameshg87: yup, image_service will use it. i've just been holding off on making changes on that patch until the spec is finished | 17:13 |
| rameshg87 | JoshNang, great .. that brings me to the second question | 17:13 |
| rameshg87 | JoshNang, i have some questions on how tmp url thing works with glance backed images .. | 17:13 |
| rameshg87 | JoshNang, the direct_url looks like this for me: swift+http://tenant%3Ausername:password@host:5000/v2.0/container/object | 17:14 |
| rameshg87 | JoshNang, i hope the glance direct urls look this for you as well ? | 17:14 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Rename/update common/tftp to drivers/modules/pxe_utils https://review.openstack.org/103595 | 17:15 |
| JoshNang | rameshg87: right. they look the same for me. and i'd like to transform that to http://host:5000/v2.0/container/object?duration=..... | 17:15 |
| *** Penick has joined #openstack-ironic | 17:15 | |
| rameshg87 | JoshNang, this generated url of the above form is not working for me | 17:16 |
| rameshg87 | JoshNang, as i understand http://host:5000/ is used by keystone | 17:16 |
| rameshg87 | JoshNang, as i understand, the request for temp urls is handled by swift proxy server which in devstack listens on http://host:8080 | 17:17 |
| *** harlowja has joined #openstack-ironic | 17:17 | |
| rameshg87 | JoshNang, when i try to do a wget on the url which looks like http://host:5000/v2.0/container/object?duration=., it's not working for me | 17:17 |
| rameshg87 | JoshNang, how does it work ? | 17:17 |
| JoshNang | rameshg87: hrm. i'm not sure on the keystone/proxying (we don't use keystone yet). swift side: here's what happens: https://github.com/openstack/swift/blob/master/swift/common/middleware/tempurl.py. when i originally wrote the patch it was working with a devstack server, but let me spin up a new one and test it | 17:20 |
| rameshg87 | JoshNang, i did some experiments. | 17:21 |
| rameshg87 | JoshNang, correct me if i am wrong | 17:21 |
| rameshg87 | JoshNang, the direct url presented by glance looks like the following: swift+http://tenant%3Ausername: password@host:5000/v2.0/container/object | 17:21 |
| NobodyCam | romcheg: -1'd 103189 for the same reason you orginally did. | 17:22 |
| rameshg87 | JoshNang, we can connect to swift in two ways as documented in client | 17:22 |
| rameshg87 | JoshNang, either using username, password, tenant_name, authurl OR using preauthurl, preauthtoken | 17:22 |
| *** jbjohnso_ has joined #openstack-ironic | 17:23 | |
| romcheg | Ah? what? what did I do? ^) | 17:23 |
| NobodyCam | romcheg: you removed your -1 of 103189... | 17:23 |
| rameshg87 | JoshNang, either way you connect, you have a thing called storage url which looks like http://host:8080/v1/AUTH_<tenant_id> | 17:23 |
| NobodyCam | romcheg: I added back a -1 for the same thing | 17:24 |
| romcheg | NobodyCam: I set it to +1 because I'm not sure that's the problem in the code | 17:24 |
| rameshg87 | JoshNang, you combine your (container, object) into /v1/AUTH_<tenant_id> of storage url to get the storage url for your object | 17:24 |
| NobodyCam | see my comment. Making htat change would also save a redirect | 17:24 |
| *** jbjohnso has quit IRC | 17:24 | |
| rameshg87 | JoshNang, so it looks like http://host:8080/v1/AUTH_<tenant_id>/<container>/<object> | 17:25 |
| JoshNang | rameshg87: ok. right | 17:26 |
| rameshg87 | JoshNang, it is from this that we generate swift temp url and then append it to your swift object url http://host:8080/v1/AUTH_<tenant_id>/<container>/<object>?duration=xxx,sig=yyyy | 17:26 |
| rameshg87 | JoshNang, the swift proxy server listening on host:8080 (or any configured port) would know how to handle such a request given the signature and duration | 17:26 |
| *** dwalleck has joined #openstack-ironic | 17:27 | |
| JoshNang | rameshg87: gotcha gotcha. i'm more and more thinking dropping support for direct_url may be the cleaner approach | 17:27 |
| rameshg87 | JoshNang, i doubt if we can fetch from a url looking like http://host:5000/v2.0/container/object?duration=..... | 17:27 |
| rameshg87 | JoshNang, oh any reason for that ? we also need temp url feature :-) | 17:28 |
| romcheg | NobodyCam: I thought one should never say the Mac or Safari word in open-source communities :) | 17:28 |
| *** dwalleck has quit IRC | 17:29 | |
| rameshg87 | JoshNang, i would be happy to help to get it in :-) | 17:29 |
| JoshNang | rameshg87: i just meant that only using a config option that specifies something to the effect of http://host:8080/v1/AUTH_<tenant_id> maybe cleaner and less error prone | 17:30 |
| rameshg87 | JoshNang, the only this is if you are given a url which looks like this "swift+http://tenant%3Ausername: password@host:5000/v2.0/container/object", you need to convert it to a url which looks like "http://host:8080/v1/AUTH_xxx/container/object" | 17:31 |
| NobodyCam | romcheg: LOL... you are probility correct | 17:31 |
| *** jbjohnso_ has quit IRC | 17:32 | |
| *** jbjohnso_ has joined #openstack-ironic | 17:32 | |
| rameshg87 | JoshNang, for doing this conversion, you can get it from keystone, you need th public_url for the swift service in keystone | 17:32 |
| rameshg87 | JoshNang, or connect to swift and make a get_auth() call | 17:32 |
| *** jbjohnso__ has joined #openstack-ironic | 17:33 | |
| rameshg87 | JoshNang, since we have all the information, i think we can do it without asking end user. | 17:33 |
| rameshg87 | JoshNang, without asking user to add a config entry | 17:34 |
| JoshNang | rameshg87: that would probably be easiest for ops/deployers | 17:34 |
| rameshg87 | JoshNang, yeah i think it's possible. i will try to take your patch and try to modify and see if we can get it. | 17:36 |
| *** jbjohnso_ has quit IRC | 17:36 | |
| NobodyCam | lucasagomes: looking at https://review.openstack.org/#/c/103136 | 17:37 |
| NobodyCam | would anything in https://github.com/openstack/ironic/blob/master/ironic/nova/scheduler/ironic_host_manager.py need to be updated along with it? | 17:38 |
| rameshg87 | NobodyCam, thanks for the ilo power driver review :-) | 17:38 |
| NobodyCam | rameshg87: Thank you for the Spec | 17:38 |
| JoshNang | rameshg87: awesome! that'd be sweet. i'll also see what i can write for it, but i probably won't have time to get to it for a couple days | 17:38 |
| rameshg87 | JoshNang, i would try to get my hands onto it. will keep you updated. | 17:39 |
| JoshNang | rameshg87: thanks! | 17:39 |
| rameshg87 | JoshNang, thanks a lot for your time. | 17:39 |
| lucasagomes | NobodyCam, not that I can see, looking at https://github.com/openstack/nova/blob/master/nova/scheduler/host_manager.py | 17:39 |
| lucasagomes | they don't see to use KBytes (or convert from it) anywhere | 17:40 |
| JoshNang | rameshg87: thanks for helping me make the patch better | 17:40 |
| *** pelix has quit IRC | 17:40 | |
| rameshg87 | devananda, can you please have a look at the ilo power spec: https://review.openstack.org/#/c/97455/ | 17:40 |
| rameshg87 | JoshNang, :-) | 17:40 |
| *** bvivek has quit IRC | 17:41 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 17:41 | |
| *** lucasagomes is now known as lucas-afk | 17:42 | |
| openstackgerrit | Divakar Padiyar Nandavar proposed a change to openstack/ironic: Fix the section name in CONTRIBUTING.rst https://review.openstack.org/103189 | 17:44 |
| *** rameshg87 has quit IRC | 17:51 | |
| * devananda prepares notes for the meeting, wanders off to find food | 18:11 | |
| jroll | jbjohnso__: hey, is there a difference between 'chassis bootdev' and 'chassis bootparam'? | 18:11 |
| * romcheg is trying to finish the boss before the meeting starts | 18:12 | |
| *** dwalleck has joined #openstack-ironic | 18:12 | |
| romcheg | NobodyCam: I figured out the problem with that URL | 18:17 |
| openstackgerrit | Divakar Padiyar Nandavar proposed a change to openstack/ironic: Fix the section name in CONTRIBUTING.rst https://review.openstack.org/103189 | 18:24 |
| *** coolsvap is now known as coolsvap|afk | 18:24 | |
| *** linggao has joined #openstack-ironic | 18:25 | |
| NobodyCam | lucas-afk: are you back by chance? | 18:29 |
| *** lsmola__ has joined #openstack-ironic | 18:31 | |
| Shrews | Is anyone else seeing py27 tests fail on the latest HEAD? | 18:32 |
| Shrews | FAIL: ironic.tests.test_glance_service.TestGlanceImageService.test_get_image_service__no_client_set_https | 18:33 |
| Shrews | FAIL: ironic.tests.test_glance_service.TestGlanceImageService.test_check_image_service__no_client_set_http | 18:33 |
| Shrews | both with -- AttributeError: 'Client' object has no attribute 'endpoint' | 18:33 |
| Shrews | I don't see how that's possible, but I've experienced it on two different machines after recreating the py27 env | 18:35 |
| *** dtantsur has quit IRC | 18:41 | |
| lucas-afk | NobodyCam, hi there | 18:42 |
| lucas-afk | yes | 18:42 |
| NobodyCam | :) | 18:42 |
| NobodyCam | hey hey lucas-afk have you tested 103136? | 18:42 |
| NobodyCam | locally? | 18:42 |
| *** dtantsur has joined #openstack-ironic | 18:43 | |
| *** Poornima has joined #openstack-ironic | 18:43 | |
| NobodyCam | https://github.com/openstack/ironic/blob/master/ironic/nova/scheduler/filters/exact_ram_filter.py line 29.. | 18:43 |
| Shrews | hrm, python-glanceclient 0.13.1 released today... i smell a problem | 18:44 |
| lucas-afk | NobodyCam, yes hmmmm I don't think I used the extact ram filter tho | 18:45 |
| jroll | Shrews: uh oh | 18:45 |
| * jroll gets ready for the meeting | 18:45 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Drivers determine acceptable power states https://review.openstack.org/86744 | 18:49 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Allow long-running ramdisks for agent driver https://review.openstack.org/103622 | 18:49 |
| *** lsmola__ has quit IRC | 18:50 | |
| *** lucas-afk is now known as lucasagomes | 18:53 | |
| *** mrda-away is now known as mrda | 18:58 | |
| mrda | Morning Ironic! | 18:58 |
| lucasagomes | devananda, NobodyCam sorry I will have to be a bit afk in the meeting :( | 18:58 |
| lucasagomes | mrda, morning | 18:58 |
| NobodyCam | morning mrda | 18:59 |
| NobodyCam | lucasagomes: its ok | 18:59 |
| mrda | \o | 18:59 |
| romcheg | Good morning mrda-away | 18:59 |
| romcheg | mrda even | 18:59 |
| *** rloo has joined #openstack-ironic | 19:00 | |
| mrda | I feel like I'm away still :) | 19:00 |
| romcheg | hehe | 19:00 |
| NobodyCam | heheh | 19:00 |
| romcheg | Someone is still in the room | 19:00 |
| Shrews | gah. jroll: https://jenkins07.openstack.org/job/gate-ironic-python26/569/console | 19:03 |
| Shrews | everything will fail until we run this down | 19:03 |
| *** lsmola__ has joined #openstack-ironic | 19:03 | |
| Shrews | jroll: that's from your 103622 job, btw | 19:04 |
| jroll | Shrews: yeah. meeting :) | 19:04 |
| Shrews | multitask!!! :) | 19:04 |
| jroll | lol | 19:05 |
| jroll | I'm also eating lunch | 19:05 |
| jroll | and potentially listening to a tech talk | 19:05 |
| jroll | this meeting time is bad for our team :( | 19:05 |
| ekarlso | what techtalk ? | 19:06 |
| JayF | at our offices (Rackspace in SF) we have office lunch + tech talk on Mondays at noon. | 19:06 |
| jroll | our office does a tech talk every monday | 19:06 |
| jroll | just internal people, not visitors or whatever | 19:06 |
| JayF | so we have that + ironic meeting every monday | 19:06 |
| *** adam_g` is now known as adam_g | 19:07 | |
| lucasagomes | NobodyCam, are u going to test with the exact filter? I won't have much time this week :( going to the ceilometer mid-cycle tomorrow | 19:07 |
| NobodyCam | lucasagomes: I will try toi | 19:08 |
| NobodyCam | to | 19:08 |
| *** lsmola__ has quit IRC | 19:08 | |
| lucasagomes | NobodyCam, ack, thanks | 19:09 |
| jbjohnso__ | jroll, nope | 19:11 |
| jbjohnso__ | (I'm verry high latency | 19:12 |
| jbjohnso__ | chassis bootparam question | 19:12 |
| soren | I'm working on getting a basic Ironic installation working and I'm struggling with the networking part. The install guide says /nothing/ about Neutron configuration, yet Neutron seems a critical component in Ironic's architecture. | 19:12 |
| soren | My problem may just be that I'm an idiot when it comes to Neutron. :) | 19:15 |
| soren | Suppose I'm on an unrestricted network. Just two physical nodes plugged into a dumb switch. | 19:16 |
| NobodyCam | soren: we are in a meeting right now. prob 45 minutes until it over | 19:17 |
| soren | Does anyone have a sample config (mostly for Neutron, but Ironic and Nova configs for completeness would be nice) that'll let me run one node as a "controller" that runs something on the other? | 19:17 |
| soren | NobodyCam: Gotcha. Thanks. | 19:17 |
| jroll | jbjohnso__: thanks | 19:23 |
| boris-42 | ooo | 19:23 |
| boris-42 | hi all | 19:23 |
| NobodyCam | hi boris-42 we are in #openstack-meeting | 19:24 |
| boris-42 | NobodyCam heh I thought I am writing in meeting lol | 19:25 |
| lucasagomes | devananda, re spec review day, I dunno if I will be around :( I'll be at https://wiki.openstack.org/wiki/Sprints/ParisJuno2014, but I will try to review/fix the specs early on wednesday | 19:27 |
| *** amitpp has joined #openstack-ironic | 19:35 | |
| devananda | lucasagomes: which spec review day? I've proposed two | 19:38 |
| lucasagomes | devananda, oh sorry, the next one July 2 | 19:39 |
| jroll | boris-42: it seems to me that code that interacts with every request should have a spec. it has *some* overhead, it runs code. maybe negligible, but still. :/ | 19:55 |
| *** ndipanov has quit IRC | 19:57 | |
| boris-42 | jroll there is no overhead... | 19:57 |
| boris-42 | jroll is "if not none" big overhead? | 19:57 |
| JayF | It is overhead at all, even if it's negligable :) | 19:58 |
| jroll | I didn't say big. I said there's some. it executes instructions on the cpu | 19:58 |
| boris-42 | JayF fully disagree with this | 19:58 |
| jroll | and knowing python, 'if not none' likely takes over 100 cycles | 19:58 |
| boris-42 | =) | 19:58 |
| boris-42 | JayF jroll guys it's nothging | 19:58 |
| jroll | boris-42: when you're running something at scale, everything needs to be considered :| | 19:59 |
| boris-42 | comparing to everything else | 19:59 |
| boris-42 | jroll nope not this | 19:59 |
| boris-42 | jroll we have so many crap | 19:59 |
| boris-42 | jroll so many unscalalble places | 19:59 |
| boris-42 | jroll that we shouldn't think about if None | 19:59 |
| boris-42 | jroll e.g. authenticate in keystone | 19:59 |
| boris-42 | will take couple of secs (or even more) | 20:00 |
| boris-42 | and compare it to if None | 20:00 |
| boris-42 | that takes couple of cycles | 20:00 |
| JayF | I'm fully in agreement with jroll that I would love to see a spec | 20:00 |
| boris-42 | jroll plus you can turn of middleware | 20:00 |
| jroll | boris-42: can I see the middleware? | 20:00 |
| boris-42 | jroll JayF guys | 20:00 |
| jroll | I'm not fighting about this | 20:01 |
| jroll | because I can't make this decision anyway | 20:01 |
| *** lucasagomes is now known as lucas-afk | 20:01 | |
| boris-42 | jroll JayF this specs | 20:01 |
| boris-42 | will kill openstack | 20:01 |
| NobodyCam | good meeting all ... Thank you | 20:01 |
| lucas-afk | alright folks have a good night everyone | 20:01 |
| devananda | linggao: i think that line item was more about makign ipminative default and removing ipmitool -- this had no support outside of IBM that I'm aware of | 20:01 |
| boris-42 | I am thinking about going to do some other projects | 20:01 |
| NobodyCam | br | 20:01 |
| boris-42 | without specs crap | 20:01 |
| NobodyCam | brb | 20:01 |
| boris-42 | cause….. it's bueeee | 20:01 |
| linggao | devananda, I see. thanks for the clarification. | 20:02 |
| devananda | linggao: I've had a few reports from folks who have started testing with pyghmi that it has issues -- I hope they are filing bugs, because I haven't gotten any specific feedback yet, just "we're OK with ipmitool" | 20:02 |
| devananda | linggao: but the ipminative driver is in tree, so that line item does seem confusing. I'll update before sending the email out | 20:02 |
| devananda | thanks | 20:02 |
| boris-42 | jroll JayF https://github.com/stackforge/osprofiler/blob/master/osprofiler/web.py | 20:02 |
| *** wanyen has joined #openstack-ironic | 20:02 | |
| JayF | jroll: did we have any trouble with pyghmi? I know we aren't running it but I didn't remember why | 20:02 |
| boris-42 | jroll JayF middleware | 20:02 |
| jroll | JayF: it was a bit unstable | 20:02 |
| mrda | devananda: so just FYI, I've got a patch in flight for credential caching (https://review.openstack.org/#/c/102695) - but currently failing check-tempest-dsvm-virtual-ironic. romcheg is helping me debug :) but it'd be nice to get it into the nova driver once we get through the gate. | 20:02 |
| boris-42 | if enabled: | 20:02 |
| jroll | JayF: losing connections etc | 20:03 |
| boris-42 | if not enabled: | 20:03 |
| linggao | devananda, sure. We'll make sure to test it. | 20:03 |
| boris-42 | will return erquest | 20:03 |
| boris-42 | no overhead | 20:03 |
| JayF | jroll: anything we can specifically bug? | 20:03 |
| jroll | JayF: not without turning it back on and poking at it | 20:03 |
| *** rloo has quit IRC | 20:05 | |
| jroll | boris-42: the question about overhead is about "what's the overhead if this is *enabled*" | 20:05 |
| adam_g | mrda, which review was showing those neutron issues? | 20:05 |
| boris-42 | jroll so it's not the big | 20:05 |
| boris-42 | jroll you will make 1 if not True | 20:05 |
| boris-42 | and then you'll do 2 headers.get("key") | 20:06 |
| devananda | mrda: ++ | 20:06 |
| boris-42 | as they will be empty you'll skip section with valid header | 20:06 |
| boris-42 | and run your method | 20:06 |
| boris-42 | jroll so overhead | 20:06 |
| boris-42 | 2 get in small dict, one trivial if | 20:07 |
| boris-42 | jroll ^ pls don't be silly … it's not thing that we should think | 20:07 |
| jroll | and... send notifications? | 20:07 |
| mrda | adam_g: This https://review.openstack.org/#/c/102695 is failing check-tempest-dsvm-virtual-ironic - with potential neutron issues, but still debugging. I'm no tempest expert so it's slow going for me to work out why it's failing. romcheg has been helpful :) | 20:07 |
| boris-42 | jroll nope | 20:07 |
| boris-42 | jroll there is no notifications | 20:07 |
| jroll | what | 20:07 |
| boris-42 | jroll if there is no special trace headers | 20:07 |
| jroll | then what is the purpose | 20:07 |
| boris-42 | jroll ? | 20:07 |
| boris-42 | jroll only if profiler is enabled | 20:08 |
| jroll | so | 20:08 |
| jroll | if I enable the profiler | 20:08 |
| boris-42 | jroll and you are sending special headers in HTTP request | 20:08 |
| jroll | I would probably like to profile things | 20:08 |
| boris-42 | jroll then you'll got your notifications | 20:08 |
| jroll | is that going to kill my control plane? | 20:08 |
| jroll | is that going to cause performance issues? | 20:08 |
| boris-42 | jroll ^ read what I wrote | 20:08 |
| jroll | I'm asking rhetorical questions | 20:08 |
| jroll | because this is what the spec is for | 20:08 |
| jroll | so those questions are answered up front | 20:09 |
| jroll | and people don't need to ask them | 20:09 |
| devananda | what is the API to enable profiler? | 20:09 |
| jroll | you've spent more time arguing about specs than it would take to write them | 20:09 |
| adam_g | mrda, hmm looks like its actually an issue with the patch: http://logs.openstack.org/95/102695/1/check/check-tempest-dsvm-virtual-ironic/78de7c1/logs/screen-n-cpu.txt.gz#_2014-06-27_06_04_21_321 | 20:09 |
| boris-42 | jroll nope | 20:09 |
| boris-42 | jroll I don't want to create 20 specs | 20:10 |
| jroll | devananda: sounds like, "add this header to your request" | 20:10 |
| devananda | can it be enabled without downtime? per request? how does that work between services? | 20:10 |
| devananda | how does taht header get forwarded from nova to ironic? | 20:10 |
| mrda | adam_g: cool! thanks. | 20:10 |
| adam_g | mrda, np :) | 20:10 |
| * jroll is apparently answering rhetorical questions now. /me walks away for a bit | 20:10 | |
| boris-42 | devananda it's quite simple lemme show you patches in cinder | 20:10 |
| devananda | boris-42: sometimes simple code has large ramifications | 20:10 |
| devananda | boris-42: that people want to undersatnd | 20:11 |
| boris-42 | devananda https://review.openstack.org/#/c/103359/2/cinderclient/client.py | 20:11 |
| devananda | boris-42: is there a document somewhere which explains all this? | 20:11 |
| boris-42 | https://github.com/stackforge/osprofiler | 20:11 |
| boris-42 | ^ readme | 20:11 |
| * mrda wanders away for an hour - going for a run before it rains again here. bbs. | 20:11 | |
| boris-42 | devananda take a look at that read me ^ it's everything about this lib | 20:11 |
| boris-42 | devananda how it works, how to integrate, and why you need it=0 | 20:12 |
| devananda | boris-42: link to patch in cinder adding support for it? | 20:12 |
| boris-42 | devananda it adds support to python client | 20:12 |
| boris-42 | devananda to add header if profiler is initalized | 20:12 |
| devananda | boris-42: right. where's the suport in cinder | 20:12 |
| jroll | actual question: it seems like it is enabled in cinder by installing osprofiler in the virtualenv? | 20:12 |
| boris-42 | devananda https://review.openstack.org/#/c/103415/ | 20:12 |
| wanyen | Deva: I have questions on some of the specs that iLO developers submitted. Based on the review comments, we have refactor specs into genreic and iLO specific. My question is wheterh a BMC interface-specific (e.g., ipmitool or iLO) spec is needed given that it is just using a specifc BMC interface to implement the generic spec. | 20:13 |
| jroll | s/cinder/cinderclient/ | 20:13 |
| boris-42 | devananda service file should be refactored cause markmc dislike idea of putting that part to oslo.messaging (but it's not so big issue) and I'll address tomororw | 20:13 |
| jroll | enabled by default. love it. | 20:13 |
| boris-42 | jroll even if it is enabled it won't produce load | 20:14 |
| *** amitpp has quit IRC | 20:14 | |
| boris-42 | jroll until it get headers | 20:14 |
| boris-42 | to get headers you need to do something like | 20:14 |
| boris-42 | # get somehow ironicclient | 20:14 |
| jroll | which are sent for every cinder client request, if osprofiler is installed next to the client? | 20:14 |
| boris-42 | profiler.init() | 20:14 |
| boris-42 | ironiccleint.do_some_action() | 20:15 |
| devananda | boris-42: the additional load on the *client* doesn't matter | 20:15 |
| boris-42 | ^ then do_some_action will add headers | 20:15 |
| devananda | boris-42: what is the impact on the *server* | 20:15 |
| jroll | so now my customer can install osprofiler and DOS my service | 20:15 |
| boris-42 | jroll nope | 20:15 |
| boris-42 | jroll there is HMAC key | 20:15 |
| *** dwalleck has quit IRC | 20:15 | |
| boris-42 | profiler.init(hmac="secret_key") | 20:15 |
| boris-42 | ^ only if secret_key is that is specified in wsgi middleware | 20:16 |
| boris-42 | this stuff will work | 20:16 |
| jroll | right, I understand that | 20:16 |
| devananda | boris-42: so I want to see a spec | 20:16 |
| boris-42 | devananda probably I can make spec in oslo? | 20:16 |
| devananda | boris-42: has this been approved for nova? | 20:16 |
| boris-42 | devananda and send everybody to it? | 20:16 |
| devananda | that's where I'd start | 20:16 |
| devananda | or oslo | 20:16 |
| boris-42 | devananda fff nova=) | 20:16 |
| boris-42 | =) | 20:16 |
| boris-42 | nova is last point | 20:16 |
| boris-42 | =) | 20:16 |
| devananda | also -- IIRC from the summit, you want to use Ceilometer as the bus for these profiling event notices | 20:16 |
| boris-42 | never start from nova | 20:17 |
| boris-42 | if you would like to get it | 20:17 |
| devananda | which several folks (myself included) had some serious concerns about | 20:17 |
| devananda | since Ceilometer is already havign scaling problems | 20:17 |
| boris-42 | devananda so lemme expalin | 20:17 |
| boris-42 | devananda I would like to get everything working | 20:17 |
| boris-42 | devananda then I'll create special backend in ceilometer | 20:18 |
| boris-42 | for stroing profiling data | 20:18 |
| boris-42 | the second direction of work is to add support of other storages | 20:18 |
| boris-42 | e.g. just plain mongo | 20:18 |
| boris-42 | or sql | 20:18 |
| boris-42 | via configurations | 20:18 |
| boris-42 | devananda but for now believe it's really hard to get anything in this openstack | 20:19 |
| boris-42 | devananda that affects all projects | 20:19 |
| boris-42 | devananda so guys I think you should try to reduce my pain=) | 20:19 |
| boris-42 | devananda step by step | 20:20 |
| boris-42 | devananda profiler based on Ceilometer can't be probably used in production | 20:20 |
| boris-42 | devananda but it can be used in gates | 20:20 |
| devananda | boris-42: no, it's not. it's hard to land features when your proposed implementation raises serious concerns, and you don't want to work with the process each team has adopted | 20:20 |
| boris-42 | devananda ok | 20:20 |
| NobodyCam | boris-42: that really is what oslo of for, code shared by all OpenStack | 20:21 |
| devananda | boris-42: no offense, but ironic's priorities are not "reduce your pain" -- it's to make a functional product. | 20:21 |
| boris-42 | devananda NobodyCam okay I won't' touch ironic | 20:21 |
| boris-42 | sorry guys | 20:21 |
| devananda | boris-42: we have a process by which new features should be introduced. claiming "this is so simple it doesn't need a spec" is a strawman | 20:21 |
| boris-42 | for spending your time | 20:21 |
| NobodyCam | not at all | 20:21 |
| devananda | boris-42: getting proper instrumentation in ironic is *part* of that product | 20:22 |
| NobodyCam | you had a real question and we worked through the answer | 20:22 |
| devananda | boris-42: but it's not a trivial matter and trying to skirt around our process for vetting new features is harmful to the project | 20:23 |
| *** Poornima has quit IRC | 20:25 | |
| soren | I'm really struggling to get Ironic running in the most basic setup imaginable. I'm having a particularly hard time working out how much of e.g. https://wiki.openstack.org/wiki/Baremetal I'm supposed to believe. | 20:27 |
| NobodyCam | soren: what error are you getting | 20:28 |
| *** overlayer has quit IRC | 20:30 | |
| soren | NobodyCam: I'm not really at a point yet where I can talk about specific errors. I'm still trying to connect the various pieces. | 20:32 |
| NobodyCam | soren: :) are you running devstack or devtest? | 20:33 |
| soren | NobodyCam: No. | 20:33 |
| NobodyCam | do you have currently running openstack env? | 20:34 |
| soren | No, this is a new one. | 20:34 |
| soren | Well... | 20:34 |
| soren | I suppose the actual answer is yes. | 20:34 |
| soren | I do have a running openstack env. | 20:34 |
| soren | But I'm setting up another one now. | 20:34 |
| soren | And it's this new one that I'm struggling with. | 20:34 |
| devananda | romcheg: ping | 20:35 |
| romcheg | devananda: pong | 20:35 |
| soren | I started out installing Ironic and Keystone. I got them working together. I created a chassis and a node and added all its ports. | 20:35 |
| soren | They racked servers, not blades, so the chassis is fake. | 20:36 |
| soren | *they're | 20:36 |
| NobodyCam | we not really using chassis atm | 20:36 |
| NobodyCam | so thats ok | 20:36 |
| soren | ok. | 20:36 |
| soren | I'm also running Glance. | 20:36 |
| soren | Then I installed Nova and configured it according to http://docs.openstack.org/developer/ironic/deploy/install-guide.html | 20:37 |
| soren | I can power the node on and off with the ironic CLI tool, btw. | 20:37 |
| soren | So that works. | 20:37 |
| soren | Now, that install guide doesn't talk about special flavors at all. | 20:38 |
| soren | ...and it doesn't talk about python-diskimagebuilder. | 20:38 |
| soren | ...or Neutron configuration. | 20:38 |
| soren | I tried just uploading a regular Ubuntu image to Glance and running it (with "nova boot [...]"). | 20:39 |
| soren | It failed with NoValidHost. "nova hypervisor-list" shows a single host (as expected) with -512 MB ram. | 20:39 |
| soren | The docs for nova-baremetal says to specify how much ram a node has when adding it with nova baremetal-node-create, but I get the impression that's deprecated now. | 20:39 |
| soren | ...but it's exceptionally difficult to tell. | 20:40 |
| NobodyCam | soren: ahh so your testing with out nova and the ironic nova driver | 20:41 |
| soren | Sorry about the unstructured brain dump here, but I'm just really not sure where to begin :) | 20:41 |
| soren | NobodyCam: I thought that's how things were done. | 20:41 |
| soren | NobodyCam: Sorry, wait. | 20:42 |
| soren | What? No. | 20:42 |
| soren | Without nova? | 20:42 |
| * soren is shocked to realise that there was an entire new level of confusion available | 20:42 | |
| NobodyCam | soren: the cli does now support set_provision_state | 20:43 |
| devananda | soren: so the wiki/Baremetal is applicable primarily to the (deprecated) nova.virt.baremetal driver | 20:43 |
| devananda | soren: not to ironic | 20:43 |
| NobodyCam | but I have not actually play with it | 20:43 |
| jroll | NobodyCam: soren said they are using nova boot...? | 20:43 |
| NobodyCam | doh, /me makes his window larger then 4 lines :-p | 20:44 |
| soren | devananda: That's what I thought. However, that page is the only place I could find out how to run python-diskiamgebuilder, for instance. | 20:44 |
| devananda | soren: there is no "python-diskimagebuilder" project | 20:44 |
| devananda | soren: it is "diskimage-builder" | 20:44 |
| soren | devananda: ...and it's also the only page that seemed to have any clues for how to tell anything how much RAM my node has. | 20:44 |
| NobodyCam | https://github.com/openstack/diskimage-builder | 20:45 |
| devananda | soren: https://github.com/openstack/diskimage-builder | 20:45 |
| soren | devananda: Sorry, my bad. | 20:45 |
| soren | devananda: Muscle memory, I suppose. | 20:45 |
| devananda | soren: a *lot* has changed since that page was written, and while the principles may still apply, the commands will probaly not work | 20:46 |
| soren | Ok, maybe a few questions will clear things up for me.. Do I still need to create special flavours in Nova for ironic? | 20:46 |
| devananda | soren: unless you were running the nova baremetal driver (which is stil lin Nova's tree, which is why I haven't taken down taht page yet) | 20:46 |
| devananda | soren: it depends :) | 20:46 |
| devananda | soren: if you use the exact-match filters (in ironic's tree) then you need to create flavours which have the same properties as your nodes | 20:47 |
| soren | devananda: How can I tell if I am? | 20:48 |
| devananda | soren: did you copy them from ironic/nova/scheduler/filters/* to your nova installation? | 20:48 |
| devananda | soren: or did you add them to your nova config file? | 20:49 |
| soren | devananda: I didn't. I'm using the Ironic Puppet module, though, which *may* have, but I don't think so. | 20:49 |
| devananda | soren: short answer is, you probably are not (and that's fine) | 20:49 |
| devananda | soren: huh. I didn't know someone had written a puppet module for Ironic yet. neat! got a link? | 20:49 |
| soren | https://github.com/stackforge/puppet-ironic :) | 20:50 |
| soren | devananda: ..and I'm trying my best to create a nova::compute::ironic class. | 20:50 |
| soren | Ok, so I'm probably not useing the exact-match filters. So I don't need to create special flavours? The default ones will happily get scheduled to run on my bare metal nodes through Ironic? | 20:52 |
| adam_g | soren, can you confirm that nova scheduler is picking up resources from ironic? ie, 'nova hypervisor-stats' | 20:52 |
| soren | adam_g: Sort of. I see the node, but its data is wrong. | 20:53 |
| adam_g | 'count' should be something similar to # of nodes in ironic | 20:53 |
| soren | The count is correct. | 20:53 |
| devananda | soren: so my puppet-fu isn't the best, but I think https://github.com/stackforge/puppet-ironic/blob/master/spec/classes/ironic_conductor_spec.rb should contain add'l packages too, like iscsiadm and so on | 20:53 |
| adam_g | soren, the data in the ironic node's properties is correct? via 'ironic node-show' | 20:54 |
| NobodyCam | brb | 20:55 |
| soren | adam_g: Let me check. | 20:57 |
| devananda | soren: nova's default scheduler will pick a node where node resources >= flavor resources (for all resources in cpu, ram, disk) | 20:57 |
| soren | devananda: How is nova supposed to know how much RAM it has? | 20:57 |
| devananda | soren: you tell it | 20:57 |
| devananda | soren: "ironic node-create ..." | 20:57 |
| soren | devananda: Where? When? How? | 20:57 |
| devananda | when you register the node with ironic | 20:58 |
| jroll | ... which our install guide doesn't cover :( | 20:58 |
| soren | devananda: Go on.. | 20:58 |
| devananda | ... right :( | 20:58 |
| devananda | one sec | 20:58 |
| *** jdob has quit IRC | 20:59 | |
| devananda | soren: this may (or may not) help - http://bit.ly/os-ironic-linuxtag-14 | 20:59 |
| devananda | I need to find time to convert that to something that slideshare understands | 20:59 |
| devananda | or make a video out of it | 21:00 |
| devananda | soren: see slide 11 | 21:00 |
| soren | Yup, just found it. | 21:00 |
| devananda | jroll: don't suppose you're volunteering to add a "usage guide" ;) | 21:01 |
| devananda | since this isn't really part of installing the service per se | 21:01 |
| jroll | devananda: not... really, no | 21:02 |
| jroll | :/ | 21:02 |
| devananda | :p | 21:03 |
| soren | Can I add these after the fact with "ironic node-update <uuid> add cpus=16 memory_mb=[...]"? | 21:04 |
| soren | Or is node-update for something else? It speaks of a "path" concept that I'm not familiar with. | 21:05 |
| matty_dubs | You know, my still-needs-revision blog post might be a halfway-decent starting point for a usage guide? | 21:05 |
| matty_dubs | (Except for all the stuff that's wrong.) | 21:06 |
| romcheg | Have to go now, it's *ss o'clock here. | 21:07 |
| romcheg | Have a good rest of the day guys | 21:07 |
| matty_dubs | I can try to devote a few spare cycles to a usage guide. | 21:07 |
| matty_dubs | See ya, romcheg! | 21:07 |
| *** davidlenwell_ is now known as davidlenwell | 21:07 | |
| *** linggao has quit IRC | 21:09 | |
| mrda | bye romcheg! | 21:19 |
| NobodyCam | night romcheg | 21:20 |
| *** jbjohnso__ has quit IRC | 21:22 | |
| devananda | soren: you can update node properties like that any time | 21:26 |
| devananda | soren: that change will propagate to nova asynchronously, so may take a minute or two to be reflected in "nova hypervisor-stats" | 21:27 |
| soren | devananda: Yeah, I just found out. Having to specify it as properties/<whatever> wasn't very obvious. | 21:28 |
| devananda | soren: noted | 21:28 |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: BUpdate image_source when rebuilding https://review.openstack.org/103665 | 21:29 |
| *** ccrouch1 has joined #openstack-ironic | 21:29 | |
| *** romcheg1 has joined #openstack-ironic | 21:30 | |
| soren | An example Neutron config would be really helpful, too. | 21:30 |
| *** Mikhail_D_ltp has quit IRC | 21:30 | |
| *** overlayer has joined #openstack-ironic | 21:31 | |
| *** christop1eraedo has joined #openstack-ironic | 21:31 | |
| *** Shrews_ has joined #openstack-ironic | 21:32 | |
| devananda | soren: pointing to devstack or tripleo doesn't count? ;) | 21:33 |
| devananda | (srsly tho, yes, that does need to be doc'd) | 21:33 |
| *** romcheg1 has left #openstack-ironic | 21:34 | |
| *** jgrimm has quit IRC | 21:35 | |
| *** Shrews has quit IRC | 21:35 | |
| *** kevinbenton has quit IRC | 21:35 | |
| *** romcheg has quit IRC | 21:36 | |
| *** christopheraedo has quit IRC | 21:36 | |
| *** ccrouch has quit IRC | 21:36 | |
| *** kevinbenton has joined #openstack-ironic | 21:36 | |
| *** matty_dubs is now known as matty_dubs|gone | 21:38 | |
| soren | devananda: I was under the impression that devstack would only let me run virtual bare metal? | 21:39 |
| devananda | soren: it defaults to that, but shouldn't be limited to it | 21:40 |
| devananda | as usual, devstack defaults to the gate environment | 21:42 |
| devananda | set IRONIC_VM_COUNT=0 to disable its addition of test VMs | 21:42 |
| *** blamar has quit IRC | 21:43 | |
| devananda | soren: actually i may have spoken too soon. the network bridge bits don't look easily configurable | 21:44 |
| soren | devananda: I don't really care. I have no particular desire to use devstack. | 21:44 |
| devananda | ack | 21:44 |
| *** chuckC has quit IRC | 21:58 | |
| pquerna | devananda: http://www.slideshare.net/devananda1/ods-havana-provisioning-bare-metal-with-open-stack <- slide 33, have you seen an updated diagram like that one for modern-ironic? | 22:06 |
| devananda | pquerna: the process is the same, though the service names are different | 22:07 |
| pquerna | YEAH | 22:08 |
| *** Penick has quit IRC | 22:10 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 22:17 |
| NobodyCam | devananda: fyi: I just added second +2 to https://review.openstack.org/#/c/103137 but did not approve. its lucas's driver patch. | 22:18 |
| GheRivero | Can I get some love for ^^ | 22:18 |
| NobodyCam | GheRivero: will that fix the gate tests | 22:18 |
| GheRivero | i hope so | 22:18 |
| NobodyCam | :-p | 22:18 |
| GheRivero | it passes the tests locally | 22:18 |
| NobodyCam | GheRivero: is there a patch that changed this in glance.. and can you ref it in the commit message? | 22:20 |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 22:22 |
| NobodyCam | TY | 22:22 |
| NobodyCam | lol I dont know wether to laugh or cry that adding a zero line file can cause check-tempest-dsvm-postgres-full to fail | 22:25 |
| NobodyCam | :-p | 22:25 |
| *** blamar has joined #openstack-ironic | 22:32 | |
| Shrews_ | GheRivero: NobodyCam: shouldn't we pin glanceclient based on that? | 22:38 |
| Shrews_ | otherwise, whether the tests fail or succeed will depend on which version of glanceclient is installed | 22:39 |
| Shrews_ | s/depend/depending | 22:39 |
| NobodyCam | Shrews_: pin as in that is the lowest version we can use to ping to the last version | 22:39 |
| NobodyCam | s/to ping/or pin/ | 22:40 |
| Shrews_ | NobodyCam: yeah, we should require glanceclient 0.13.1 (or whatever version that change was introduced in) | 22:40 |
| wanyen | Deva: I have questions on some of the specs that iLO developers submitted. Based on the review comments, we have refactored specs into genreic and iLO specific. My question is whether a BMC interface-specific (e.g., ipmitool or iLO) spec is needed given that it is just using a specifc BMC interface to implement the generic spec. | 22:40 |
| NobodyCam | ack good catch... | 22:40 |
| Shrews_ | ghe's fix won't work on older clients, right | 22:40 |
| NobodyCam | GheRivero: ^^^ | 22:40 |
| NobodyCam | Shrews_: comments make sense to /me | 22:41 |
| GheRivero | +1 | 22:41 |
| GheRivero | i'll work on that | 22:42 |
| NobodyCam | brb quick walkies | 22:42 |
| NobodyCam | :) ty GheRivero | 22:42 |
| pquerna | devananda: https://www.evernote.com/shard/s184/sh/041ab958-2ce7-477b-97e5-f504e51c0362/27b115a020543253a631ed3b95881956 <- roughly right? | 22:42 |
| GheRivero | Shrews_: pinning an specific version on requirements will conflict with global requirements. The code should be adapted to support both ways | 22:44 |
| Shrews_ | GheRivero: ++ | 22:45 |
| NobodyCam | GheRivero: really. :-p | 22:49 |
| *** hemna__ has joined #openstack-ironic | 22:54 | |
| *** hemna_ has quit IRC | 22:55 | |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Ironic nova driver to cache ironic client calls https://review.openstack.org/102695 | 23:09 |
| devananda | pquerna: to be completely accurate, change "nova cmpute" to "client" and first line from "driver_spawn()" to "POST /v1/nodes/NNN {provision_state: active}" | 23:10 |
| devananda | pquerna: the nova compute host running the nova.virt.ironic driver is one instance of a client | 23:11 |
| devananda | pquerna: also note that this is specific to the PXE driver. the proposed IPA and iLO drivers have different flow | 23:11 |
| openstackgerrit | Adam Gandelman proposed a change to openstack/ironic: Update Nova's available resources at termination https://review.openstack.org/102373 | 23:13 |
| adam_g | devananda, so https://review.openstack.org/103227 and https://review.openstack.org/#/c/102373/ are two from me to ironic.nova*. the rebuild one is a bit more critical, the other will be needed for the on going tempest work | 23:15 |
| devananda | adam_g: ack. i'm starting to track these on the etherpad | 23:16 |
| adam_g | devananda, cool | 23:17 |
| *** chuckC has joined #openstack-ironic | 23:17 | |
| *** openstackgerrit has quit IRC | 23:19 | |
| *** openstackgerrit has joined #openstack-ironic | 23:21 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix tear_down a node with missing info https://review.openstack.org/103685 | 23:22 |
| NobodyCam | any one know if there is a reason to set boot-file per node? I'm looking at line 148 of https://review.openstack.org/#/c/103595/1/ironic/drivers/modules/pxe_utils.py | 23:28 |
| devananda | NobodyCam: if one node uses ipxe and another does not | 23:33 |
| NobodyCam | actually we are going to cahin load ipxe so they will still need pxelinux.0 :-p from what I can tell | 23:35 |
| devananda | NobodyCam: also, we need to tell neutron the boot file name, so that it can pass that to the pxe loader on the node | 23:35 |
| devananda | NobodyCam: we need L148 regardless of whether it's pxe or ipxe | 23:35 |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix glance endpoint tests https://review.openstack.org/103670 | 23:35 |
| NobodyCam | devananda: yes, but looking at line 120 of https://review.openstack.org/#/c/103595/1/ironic/common/tftp.py (from the same review) | 23:36 |
| NobodyCam | it is a passed in var where lucas's change pins to the conf value | 23:36 |
| NobodyCam | and just get it from the configuration | 23:37 |
| NobodyCam | because we were always using the value from the configuration anyway | 23:37 |
| devananda | right | 23:38 |
| openstackgerrit | Adam Gandelman proposed a change to openstack/ironic: Update Nova's available resources at termination https://review.openstack.org/102373 | 23:38 |
| NobodyCam | is in the commit, just trying to think if there is a case where we need to support both | 23:38 |
| NobodyCam | change it per node | 23:38 |
| *** zdiN0bot1 has joined #openstack-ironic | 23:40 | |
| pquerna | devananda: thanks. yeah, i'm going to make a separate one for the IPA flow. | 23:41 |
| GheRivero | adam_g: ping | 23:41 |
| adam_g | GheRivero, hey | 23:42 |
| GheRivero | I'm just testing your patch. Does it work? I guess so, but wanted to be sure before doing it | 23:43 |
| adam_g | GheRivero, yeah, works fine. im running thru overcloud upgrade tests now on one of our racks. that patch fixes re-imaging , need to test the preserve_ephemeral bit now, which i somehow managed to not set in my last run, but it should be good | 23:45 |
| adam_g | GheRivero, re-imaging works fine /w tripleo too (when the old image has been deleted from glance prior to rebuild) | 23:46 |
| GheRivero | cool :) | 23:46 |
| *** overlayer has quit IRC | 23:50 | |
| *** zdiN0bot1 has quit IRC | 23:51 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!