| *** stevehuang has quit IRC | 00:00 | |
| *** newell_ has quit IRC | 00:00 | |
| devananda | jroll: looks like https://review.openstack.org/#/c/81919/ is failing now | 00:03 |
|---|---|---|
| JoshNang | devananda: looks like the mem bug, at least for py26 | 00:07 |
| devananda | ugh | 00:10 |
| jroll | I agree | 00:14 |
| *** rloo has quit IRC | 00:18 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add configdrive generation to Nova driver https://review.openstack.org/86192 | 00:21 |
| *** matsuhashi has joined #openstack-ironic | 00:31 | |
| *** killer_prince has quit IRC | 00:52 | |
| *** derekh has quit IRC | 00:53 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Change driver URL from teeth to agent https://review.openstack.org/86201 | 00:53 |
| *** matsuhashi has quit IRC | 00:56 | |
| *** lazy_prince has joined #openstack-ironic | 00:59 | |
| *** lazy_prince is now known as killer_prince | 00:59 | |
| *** matsuhashi has joined #openstack-ironic | 01:05 | |
| *** hemna_ has quit IRC | 01:08 | |
| *** bigjools has quit IRC | 01:11 | |
| NobodyCam | should a change like this https://review.openstack.org/#/c/85529/4/ironic/openstack/common/log.py not give debug logs in the gate jobs | 01:12 |
| *** bigjools has joined #openstack-ironic | 01:13 | |
| *** bigjools has joined #openstack-ironic | 01:13 | |
| *** ilives has joined #openstack-ironic | 01:24 | |
| *** eghobo has quit IRC | 01:28 | |
| *** nosnos has joined #openstack-ironic | 01:29 | |
| *** datajerk has joined #openstack-ironic | 01:29 | |
| *** datajerk has quit IRC | 01:33 | |
| *** datajerk has joined #openstack-ironic | 01:38 | |
| *** jgrimm has quit IRC | 02:06 | |
| *** jgrimm has joined #openstack-ironic | 02:18 | |
| *** coolsvap has joined #openstack-ironic | 02:33 | |
| *** lazy_prince has joined #openstack-ironic | 02:44 | |
| *** datajerk has quit IRC | 02:57 | |
| *** matsuhashi has quit IRC | 03:12 | |
| *** harlowja is now known as harlowja_away | 03:12 | |
| *** coolsvap has quit IRC | 03:12 | |
| *** lazy_prince has quit IRC | 03:14 | |
| *** eghobo has joined #openstack-ironic | 03:19 | |
| *** nosnos has quit IRC | 03:26 | |
| *** coolsvap1 has joined #openstack-ironic | 03:27 | |
| *** coolsvap1 has quit IRC | 03:27 | |
| *** coolsvap has joined #openstack-ironic | 03:28 | |
| *** datajerk has joined #openstack-ironic | 03:58 | |
| *** lazy_prince has joined #openstack-ironic | 04:00 | |
| *** sseago_ has quit IRC | 04:21 | |
| *** matsuhashi has joined #openstack-ironic | 04:24 | |
| *** nosnos has joined #openstack-ironic | 04:25 | |
| *** sabah has joined #openstack-ironic | 04:33 | |
| *** matsuhashi has quit IRC | 04:42 | |
| *** ironman_ has quit IRC | 04:47 | |
| *** matsuhashi has joined #openstack-ironic | 04:50 | |
| *** lazy_prince has quit IRC | 05:01 | |
| *** sabah has quit IRC | 05:25 | |
| *** datajerk has quit IRC | 05:28 | |
| *** vkozhukalov has joined #openstack-ironic | 05:39 | |
| *** pradipta_away is now known as pradipta | 05:51 | |
| openstackgerrit | Jenkins proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/83956 | 06:07 |
| *** sirushti has quit IRC | 06:08 | |
| *** _sirushti has joined #openstack-ironic | 06:09 | |
| *** _sirushti is now known as sirushti | 06:09 | |
| *** killer_prince has quit IRC | 06:13 | |
| *** lazy_prince has joined #openstack-ironic | 06:13 | |
| *** lazy_prince is now known as killer_prince | 06:13 | |
| *** max_lobur has joined #openstack-ironic | 06:17 | |
| *** eghobo has quit IRC | 06:22 | |
| *** adam_g has quit IRC | 06:41 | |
| *** max_lobur has quit IRC | 06:43 | |
| *** adam_g has joined #openstack-ironic | 06:58 | |
| *** adam_g has quit IRC | 06:58 | |
| *** adam_g has joined #openstack-ironic | 06:58 | |
| *** yuriyz has joined #openstack-ironic | 07:03 | |
| *** yuriyz has quit IRC | 07:06 | |
| *** mrda is now known as mrda_away | 07:18 | |
| *** yuriyz has joined #openstack-ironic | 07:28 | |
| *** datajerk has joined #openstack-ironic | 07:29 | |
| *** matsuhashi has quit IRC | 07:29 | |
| *** coolsvap is now known as coolsvap_away | 07:30 | |
| *** matsuhashi has joined #openstack-ironic | 07:31 | |
| *** sabah has joined #openstack-ironic | 07:32 | |
| *** datajerk has quit IRC | 07:33 | |
| dtantsur | morning Ironic | 07:34 |
| *** matsuhashi has quit IRC | 07:35 | |
| *** coolsvap_away is now known as coolsvap | 07:41 | |
| *** matsuhashi has joined #openstack-ironic | 07:49 | |
| Mikhail_D_wk | Morning dtantsur :) | 07:57 |
| Mikhail_D_wk | Morning all! :) | 07:57 |
| *** datajerk has joined #openstack-ironic | 08:02 | |
| *** jistr has joined #openstack-ironic | 08:02 | |
| *** sabah has quit IRC | 08:06 | |
| *** datajerk has quit IRC | 08:06 | |
| *** sseago has joined #openstack-ironic | 08:06 | |
| *** romcheg has joined #openstack-ironic | 08:14 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 08:23 | |
| *** rameshg87 has joined #openstack-ironic | 08:23 | |
| *** rushiagr has joined #openstack-ironic | 08:23 | |
| rushiagr | Hello folks | 08:24 |
| rushiagr | I was trying to setup Ironic with Devstack using the docs | 08:24 |
| rushiagr | but I wasn't able to .. | 08:24 |
| rushiagr | here are the devstack logs | 08:25 |
| rushiagr | http://paste.openstack.org/show/75390/ | 08:25 |
| rushiagr | and here is the error from ir-cond screen | 08:26 |
| rushiagr | http://paste.openstack.org/show/75391/ | 08:26 |
| rushiagr | strangely, errors at both the places seem unrelated | 08:26 |
| *** datajerk has joined #openstack-ironic | 08:26 | |
| *** Mikhail_D_ltp has quit IRC | 08:29 | |
| dtantsur | http://paste.openstack.org/show/75391/ <-- it's ok, error is in the first and I have never seen it :( | 08:29 |
| *** datajerk has quit IRC | 08:30 | |
| yuriyz | morning Ironic | 08:32 |
| *** lucasagomes has joined #openstack-ironic | 08:34 | |
| rameshg87 | rushiagr: i just checked the code of seamicro | 08:34 |
| yuriyz | rushiagr, there is a fix already https://review.openstack.org/#/c/85388/ | 08:34 |
| rameshg87 | rushiagr: it might be because this might be missing on your system https://github.com/seamicro/python-seamicroclient | 08:34 |
| yuriyz | but this only for seamicro drivers now | 08:35 |
| rameshg87 | rushiagr: if you are not going to use the seamicro driver, you can ignore that error as well :-) | 08:35 |
| yuriyz | yes, all others can work :) | 08:36 |
| rushiagr | I just wanted to use ironic apis with devstack. I don't want to use any drivers :) | 08:36 |
| rameshg87 | rushiagr: then you can go ahead and ignore the error :-) | 08:36 |
| *** martyntaylor has joined #openstack-ironic | 08:36 | |
| rushiagr | but it seemed like the stack.sh script exited out abruptly.. | 08:37 |
| rushiagr | it said ' Failed to define network from /dev/fd/63" | 08:38 |
| rushiagr | "XML error: unexpected virtualport type -1" | 08:38 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added lvm partitioner https://review.openstack.org/85672 | 08:40 |
| dtantsur | lucasagomes, hi. Re: https://review.openstack.org/#/c/84396/4 weren't you expected to add some tests there? Sorry if I got something wrong :) | 08:40 |
| lucasagomes | dtantsur, yeah, well... I still dunno weather that tests should go there or not (it was basically requested me to test parted other than our code itself) | 08:41 |
| lucasagomes | dtantsur, yesterday I was trying to push a milestone-proposed patch for that change | 08:42 |
| lucasagomes | and ended up updating the old patch | 08:42 |
| lucasagomes | cause we were experimenting how to push to the mp branch, and we didn't know hw to do it | 08:42 |
| dtantsur | lucasagomes, I am not sure either. The code is quite simple and I expect gparted to work... | 08:42 |
| lucasagomes | dtantsur, yeah | 08:43 |
| lucasagomes | same here | 08:43 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added lvm partitioner https://review.openstack.org/85672 | 08:43 |
| dtantsur | anyway for future changes (like in your other patches) it would be nice to have such tests, I don't remember whether you have them | 08:44 |
| lucasagomes | dtantsur, no I don't have tests partitioning a file | 08:45 |
| lucasagomes | just checking whether the commands that parted will execute are the expected commands | 08:45 |
| lucasagomes | things like that | 08:45 |
| dtantsur | lucasagomes, I'd recommend you add them, so that we don't break partitioning at some point | 08:46 |
| lucasagomes | dtantsur, ack, there's more to consider as well | 08:46 |
| dtantsur | it would be helpful e.g. to test against new versions of gparted | 08:46 |
| lucasagomes | like creating a sparse file before partitioning | 08:46 |
| lucasagomes | is fs dependent | 08:46 |
| lucasagomes | things like that | 08:46 |
| dtantsur | lucasagomes, I thing we can just skip the test if we can't create a sparse file | 08:47 |
| lucasagomes | yeah that's a valid conditional | 08:47 |
| dtantsur | most Linux fs support it somehow anyway | 08:47 |
| lucasagomes | yup | 08:47 |
| lucasagomes | I will give it a go | 08:47 |
| dtantsur | should I put these considerations as review comments? | 08:48 |
| lucasagomes | hmm sure thanks | 08:49 |
| agordeev | rushiagr: hey hey, just update libvirt to 1.1.x+ to get rid of "unexpected virtualport" | 08:51 |
| agordeev | morning Ironic | 08:51 |
| *** coolsvap is now known as coolsvap|afk | 08:52 | |
| agordeev | rushiagr: < https://review.openstack.org/#/c/70348/18/tools/ironic/templates/brbm.xml | 08:53 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk partitioner https://review.openstack.org/86163 | 09:01 |
| *** datajerk has joined #openstack-ironic | 09:02 | |
| *** coolsvap|afk is now known as coolsvap | 09:05 | |
| *** datajerk has quit IRC | 09:06 | |
| *** max_lobur has joined #openstack-ironic | 09:07 | |
| *** overlayer has joined #openstack-ironic | 09:09 | |
| *** max_lobur1 has joined #openstack-ironic | 09:09 | |
| dtantsur | agordeev, morning | 09:12 |
| *** max_lobur has quit IRC | 09:13 | |
| dtantsur | Guys, can we have some review progress with https://review.openstack.org/#/c/84497/ ? I think this one is _really_ critical. | 09:15 |
| agordeev | dtantsur: afair, it's good practice to add unittests for what being fixed. Would you like to add them? | 09:22 |
| Mikhail_D_wk | agordeev: Morning :) | 09:22 |
| agordeev | dtantsur, Mikhail_D_wk morning guys! | 09:22 |
| dtantsur | agordeev, well... It makes sense, not sure it's easy/doable. lucasagomes ^^^ ? | 09:23 |
| dtantsur | (I can try adding if you don't have time for it) | 09:23 |
| *** datajerk has joined #openstack-ironic | 09:24 | |
| agordeev | dtantsur: the reason is when the bug had been fixed and you add unittests that checks that bug was truly eliminated and won't come back in the future patches | 09:25 |
| lucasagomes | agordeev, dtantsur yeah makes sense to add tests | 09:25 |
| lucasagomes | before we couldn't run tests on the driver in the ironic tree | 09:26 |
| lucasagomes | but we merged a patch this week that allows it | 09:26 |
| dtantsur | lucasagomes, will you? If you don't have time I can try (not sure how easy it is though...) | 09:26 |
| lucasagomes | dtantsur, if u want to give it a go, np to me | 09:26 |
| lucasagomes | dtantsur, u can add ur self as co-author there | 09:27 |
| dtantsur | lucasagomes, ok, let me try :) | 09:27 |
| lucasagomes | ack | 09:27 |
| *** datajerk has quit IRC | 09:28 | |
| romcheg | Morning lucasagomes agordeev dtantsur Mikhail_D_wk! | 09:30 |
| lucasagomes | romcheg, morning | 09:30 |
| dtantsur | mordred, romcheg | 09:30 |
| dtantsur | * morning Oo | 09:30 |
| dtantsur | that was a strange misprint | 09:30 |
| romcheg | I always hit tab when writing a first word :) | 09:31 |
| dtantsur | :) | 09:31 |
| *** dtantsur is now known as dtantsur|lunch | 09:31 | |
| agordeev | romcheg: morning | 09:33 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: IPMITool to use the new ManagementInterface https://review.openstack.org/86092 | 09:37 |
| *** athomas has joined #openstack-ironic | 09:37 | |
| *** lucasagomes_ has joined #openstack-ironic | 09:43 | |
| *** matsuhas_ has joined #openstack-ironic | 09:45 | |
| *** lucasagomes has quit IRC | 09:45 | |
| *** matsuhashi has quit IRC | 09:45 | |
| *** matsuhas_ has quit IRC | 09:45 | |
| *** matsuhashi has joined #openstack-ironic | 09:46 | |
| *** datajerk has joined #openstack-ironic | 10:02 | |
| *** datajerk has quit IRC | 10:07 | |
| *** martyntaylor has quit IRC | 10:11 | |
| *** matsuhashi has quit IRC | 10:12 | |
| *** martyntaylor has joined #openstack-ironic | 10:13 | |
| *** athomas has quit IRC | 10:15 | |
| *** matsuhashi has joined #openstack-ironic | 10:16 | |
| *** athomas has joined #openstack-ironic | 10:21 | |
| *** datajerk has joined #openstack-ironic | 10:22 | |
| lucasagomes_ | GheRivero, ping re: 86063 | 10:26 |
| *** datajerk has quit IRC | 10:26 | |
| lucasagomes_ | GheRivero, I was thinking about adding a file like common/states.py to the boot devices | 10:26 |
| lucasagomes_ | and then one method to the interface to get a list of supported boot devices for that giving node | 10:26 |
| *** lucasagomes_ is now known as lucasagomes | 10:27 | |
| lucasagomes | so if drivers has a new boot device not listed there, they can add it to the mapping (common/boot_devices.py) and return it in the supported lists | 10:27 |
| GheRivero | lucasagomes: sounds good. I like it. is the 'sanitazion' also goind there? (pxe vs net) | 10:30 |
| lucasagomes | GheRivero, yeah, so we can have like boot_devices.DISK | 10:30 |
| lucasagomes | and in the IPMI driver boot_devices.DISK is mapped internally to 'disk' | 10:31 |
| lucasagomes | and the seamicro is mapped to 'hd0' | 10:31 |
| lucasagomes | for e.g | 10:31 |
| GheRivero | +1 | 10:31 |
| lucasagomes | GheRivero, ack, thanks for the comment there I will add another comment and update the patch | 10:31 |
| *** derekh has joined #openstack-ironic | 10:32 | |
| *** coolsvap is now known as coolsvap|afk | 10:38 | |
| *** eloib_ has quit IRC | 10:39 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add ManagementInterface https://review.openstack.org/86063 | 10:49 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: IPMITool to use the new ManagementInterface https://review.openstack.org/86092 | 10:49 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add ManagementInterface https://review.openstack.org/86063 | 10:52 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: IPMITool to use the new ManagementInterface https://review.openstack.org/86092 | 10:52 |
| *** matsuhashi has quit IRC | 11:05 | |
| *** matsuhashi has joined #openstack-ironic | 11:05 | |
| *** sabah has joined #openstack-ironic | 11:08 | |
| *** soren has joined #openstack-ironic | 11:12 | |
| *** sabah has quit IRC | 11:12 | |
| *** dtantsur|lunch is now known as dtantsur | 11:13 | |
| *** matsuhashi has quit IRC | 11:19 | |
| *** datajerk has joined #openstack-ironic | 11:20 | |
| *** datajerk has quit IRC | 11:24 | |
| rushiagr | agordeev: thanks. libvirt was not installed actually. IDK why :/ | 11:27 |
| *** lucasagomes is now known as lucas-hungry | 11:32 | |
| *** matsuhashi has joined #openstack-ironic | 11:36 | |
| *** matsuhashi has quit IRC | 11:39 | |
| *** pradipta is now known as pradipta_away | 11:51 | |
| *** pradipta_away is now known as pradipta | 11:53 | |
| *** pradipta is now known as pradipta_away | 11:54 | |
| *** rameshg87 has left #openstack-ironic | 11:54 | |
| *** yuriyz has quit IRC | 11:55 | |
| *** Alexei_987 has joined #openstack-ironic | 11:56 | |
| *** coolsvap|afk is now known as coolsvap | 11:57 | |
| *** datajerk has joined #openstack-ironic | 12:02 | |
| *** yuriyz has joined #openstack-ironic | 12:03 | |
| *** datajerk has quit IRC | 12:06 | |
| *** datajerk has joined #openstack-ironic | 12:18 | |
| *** datajerk has quit IRC | 12:23 | |
| NobodyCam | good morning Ironic From roswell | 12:38 |
| agordeev | NobodyCam: morning | 12:38 |
| NobodyCam | :) morning agordeev :) | 12:39 |
| Shrews | NobodyCam: morning. where to next? | 12:41 |
| NobodyCam | morning Shrews here for the day. actually going to take a bit of time today to explor roswell, Going to board the kids (dogs) for a few hours and do the rosewell tourist stuff | 12:44 |
| *** linggao has joined #openstack-ironic | 12:45 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Implement caching for master images https://review.openstack.org/85387 | 12:45 |
| NobodyCam | http://www.tripadvisor.com/Attractions-g47182-Activities-Roswell_New_Mexico.html | 12:46 |
| *** jdob has joined #openstack-ironic | 12:50 | |
| NobodyCam | but my question is why is this change not giving me debug logs in the gate jobs https://review.openstack.org/#/c/85529/4/ironic/openstack/common/log.py | 12:53 |
| lucas-hungry | morning NobodyCam | 12:53 |
| *** lucas-hungry is now known as lucasagomes | 12:53 | |
| NobodyCam | morning lucasagomes :) | 12:53 |
| *** rloo has joined #openstack-ironic | 12:58 | |
| yuriyz | morning NobodyCam | 12:58 |
| NobodyCam | morning yuriyz :) | 12:58 |
| NobodyCam | oh brb coffee is ready and quick morning walkies | 12:59 |
| *** jbjohnso_ has joined #openstack-ironic | 13:02 | |
| *** datajerk has joined #openstack-ironic | 13:02 | |
| *** rloo has quit IRC | 13:03 | |
| *** rloo has joined #openstack-ironic | 13:03 | |
| NobodyCam | wow that was quick... hehehe its cold out side | 13:04 |
| agordeev | jroll: JayF hey. I'm trying to move execute_command into utils and hitting the import loop issue again. https://review.openstack.org/#/c/85412/3/ironic_python_agent/flow.py The circular import happens among errors, base, utils | 13:05 |
| NobodyCam | lucasagomes: yuriyz: can you think of why I would not be getting debug logs with the change I pasted above? | 13:06 |
| *** datajerk has quit IRC | 13:07 | |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Add FlowExtension https://review.openstack.org/85412 | 13:07 |
| lucasagomes | NobodyCam, hmmm in the ironic.conf debug = True?! | 13:09 |
| lucasagomes | maybe some function is setting it to false in the config? | 13:09 |
| NobodyCam | conf should be comming from https://github.com/openstack/tripleo-image-elements/blob/master/elements/ironic/os-apply-config/etc/ironic/ironic.conf | 13:10 |
| NobodyCam | which has no debug in it | 13:10 |
| lucasagomes | hmmm | 13:11 |
| lucasagomes | odd | 13:11 |
| NobodyCam | oddly this: https://review.openstack.org/#/c/85101/2/elements/ironic/os-apply-config/etc/ironic/ironic.conf will give debug logs | 13:11 |
| NobodyCam | however I can set a dep on a cross project patch | 13:11 |
| NobodyCam | s/can/can not/ | 13:12 |
| * NobodyCam needs to drink the coffee | 13:12 | |
| lucasagomes | yeah it really sucks to not be able to do that | 13:13 |
| lucasagomes | but thinking idk why that doesn't give u debug logs | 13:13 |
| lucasagomes | it should | 13:13 |
| lucasagomes | gotta dig more into it to see if I can find something | 13:13 |
| lucasagomes | but looking at it, looks alright | 13:13 |
| *** nosnos has quit IRC | 13:14 | |
| NobodyCam | ya this is what I am getting in the gate logs: http://paste.openstack.org/show/IU0VfTamSMxbRuWSB3Wp/ | 13:15 |
| *** coolsvap is now known as coolsvap|afk | 13:18 | |
| *** coolsvap|afk is now known as coolsvap | 13:20 | |
| *** coolsvap is now known as coolsvap|afk | 13:22 | |
| lucasagomes | yeah no debug | 13:23 |
| *** matty_dubs|gone is now known as matty_dubs | 13:26 | |
| NobodyCam | morning matty_dubs :) | 13:27 |
| NobodyCam | node looks like it added correctly : http://logs.openstack.org/29/85529/4/check-tripleo/check-tripleo-ironic-undercloud-precise/8b381ba/console.html#_2014-04-09_00_11_52_935 | 13:30 |
| *** killer_prince has quit IRC | 13:34 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: SeaMicro to use the new ManagementInterface https://review.openstack.org/86328 | 13:42 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 13:46 |
| NobodyCam | ok really hacky attempt to get debug logs | 13:46 |
| lucasagomes | NobodyCam, maybe it's better to set a -2 on that review? so it would prevent it from being update | 13:47 |
| lucasagomes | being merged* | 13:48 |
| lucasagomes | but you still could send new patch-sets to it | 13:48 |
| NobodyCam | not sure -2 will get tested.. WIP ok? | 13:48 |
| lucasagomes | NobodyCam, sure | 13:48 |
| lucasagomes | hmm idk as well lemme check the -2 ting | 13:49 |
| lucasagomes | thing | 13:49 |
| NobodyCam | wip set | 13:49 |
| lucasagomes | NobodyCam, -2 still testing it | 13:50 |
| lucasagomes | https://review.openstack.org/#/c/78787/ | 13:50 |
| lucasagomes | I saw it when deva -2'd that patch | 13:50 |
| lucasagomes | then I sent a new review and tests were trigged | 13:50 |
| lucasagomes | but anyway, wip is good enough | 13:50 |
| NobodyCam | :) | 13:50 |
| NobodyCam | I'm watching it in zuul | 13:51 |
| agordeev | JayF: jroll nevermind, i'd found the solution | 13:56 |
| matty_dubs | Oh, a belated good morning, NobodyCam -- I fail at IRC today | 13:57 |
| NobodyCam | its all good matty_dubs :) | 13:57 |
| *** datajerk has joined #openstack-ironic | 14:02 | |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Resolve circular import, introduce ExecuteCommandMixin https://review.openstack.org/86341 | 14:16 |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Add FlowExtension https://review.openstack.org/85412 | 14:16 |
| NobodyCam | lucasagomes: wow no test cheanges needed for 84396 | 14:20 |
| lucasagomes | NobodyCam, yeah, sfdisk is not being tested in trunk heh | 14:21 |
| lucasagomes | NobodyCam, usually the tests just mock the make_partitions func | 14:21 |
| lucasagomes | I can add some tests to that | 14:22 |
| lucasagomes | but urgh, I'm thinkin about the tests on a sparse file, and it sucks | 14:22 |
| lucasagomes | what will sucks more will be to parse the output to read the part table | 14:22 |
| lucasagomes | -.- | 14:22 |
| NobodyCam | i would have thought we should at least test that it is called | 14:23 |
| lucasagomes | NobodyCam, +1 | 14:23 |
| NobodyCam | >.< | 14:23 |
| lucasagomes | lemme add some tests to it | 14:24 |
| NobodyCam | want to add a test to just check that it is called? and I'll +2 | 14:24 |
| NobodyCam | :) | 14:24 |
| lucasagomes | NobodyCam, will do | 14:24 |
| NobodyCam | as it is going to be back ported maight be worth it :) | 14:24 |
| NobodyCam | *might | 14:24 |
| lucasagomes | +1 I will update both reviews | 14:25 |
| NobodyCam | you da man :) :-p | 14:25 |
| *** saju_m has joined #openstack-ironic | 14:26 | |
| lucasagomes | NobodyCam, :D, btw so ur going to have ur RV in atlanta? I really want to see it if possible | 14:27 |
| NobodyCam | sure :) | 14:27 |
| NobodyCam | and meet the kids :) | 14:28 |
| matty_dubs | Ooh, nice! You'll have to organize a tour group or something. ;) | 14:29 |
| NobodyCam | LOL :) | 14:29 |
| lucasagomes | NobodyCam, hah awesome! | 14:29 |
| lucasagomes | lol | 14:29 |
| lucasagomes | matty_dubs, +1 | 14:29 |
| devananda | morning, all | 14:32 |
| lucasagomes | morning devananda | 14:32 |
| romcheg | Morning guys! | 14:33 |
| dtantsur | devananda, morning | 14:33 |
| NobodyCam | morning romcheg :) | 14:33 |
| NobodyCam | mornig devananda :) | 14:33 |
| NobodyCam | morning dtantsur | 14:34 |
| agordeev | devananda: morning | 14:34 |
| GheRivero | morning Ironic! | 14:36 |
| NobodyCam | morning GheRivero | 14:36 |
| devananda | lucasagomes: adding tests to 84396 and 86179 ? | 14:36 |
| lucasagomes | devananda, yeah | 14:36 |
| lucasagomes | doing it right now | 14:36 |
| devananda | awesome | 14:38 |
| devananda | rloo: hi! I just noticed https://review.openstack.org/#/c/84545/2 and that it didn't "build" into the claimed location | 14:43 |
| devananda | rloo: this is because python-ironicclient does not have a doc build hook in -infra | 14:43 |
| rloo | devananda: #@%#. I didn't even notice! | 14:44 |
| rloo | devananda: what does it mean, it "doesn't have a doc build hook in -infra"? | 14:44 |
| devananda | rloo: he requested URL /developer/python-ironicclient/ was not found on this server | 14:44 |
| devananda | woops | 14:44 |
| rloo | devananda: how did you notice it? the review it self says it merged? | 14:45 |
| devananda | yea, but click the link | 14:45 |
| devananda | it's not there | 14:45 |
| devananda | and jenkins tests dont show a doc build tset | 14:45 |
| devananda | test | 14:45 |
| rloo | ha ha. (sort of.) | 14:45 |
| devananda | rloo: is there a precedent for other projects' clients to have a similar and separate doc page? | 14:45 |
| rloo | yes | 14:46 |
| *** jgrimm has quit IRC | 14:46 | |
| devananda | great | 14:46 |
| rloo | eg: glance http://docs.openstack.org/developer/python-glanceclient/ | 14:46 |
| devananda | here's what's missing | 14:46 |
| devananda | http://git.openstack.org/cgit/openstack/ironic/tree/setup.cfg#n47 | 14:46 |
| rloo | it's missing an upload_sphinx? | 14:47 |
| *** coolsvap|afk is now known as coolsvap | 14:48 | |
| NobodyCam | bbt.. brb | 14:48 |
| devananda | http://git.openstack.org/cgit/openstack-infra/config/tree/modules/openstack_project/files/zuul/layout.yaml#n1266 | 14:48 |
| devananda | http://git.openstack.org/cgit/openstack-infra/config/tree/modules/openstack_project/files/zuul/layout.yaml#n769 | 14:49 |
| devananda | rloo: specifically, around L1266, it needs a openstack-client-publish-jobs | 14:50 |
| rloo | devananda: ok, in a template section. what about the other stuff in template: that other projects have? | 14:51 |
| devananda | rloo: sorry, still waitingn for coffee to kick in -- it seems to have the build_sphinx, but it isn't getting called, and i suspect it's the missing openstack-client-publish-jobs | 14:52 |
| devananda | rloo: the good folks in -infra know all and tell all :) | 14:52 |
| linggao | matty_dubs and NobodyCam, good morning. I found where the problem was. I added an extra pair of quotes when passing the ipmi command to shellinaboxd. | 14:52 |
| rloo | devananda: I notice that python-glanceclient has an 'upload_sphinx' in their setup.cfg. | 14:52 |
| rloo | devananda: I'll poke around and check with -infra. | 14:53 |
| linggao | matty_dubs and NobodyCam, that's why it worked manually but not in the subproces. | 14:53 |
| devananda | rloo: thanks! also, thanks! | 14:53 |
| rloo | devananda: thank YOU for noticing ;) | 14:54 |
| NobodyCam | linggao: ahh doh | 14:54 |
| NobodyCam | :) | 14:54 |
| NobodyCam | thats always tuff to spot | 14:54 |
| linggao | NobodyCam, yes. :( | 14:56 |
| *** saju_m has quit IRC | 14:58 | |
| matty_dubs | linggao: Ah, subtle! | 14:59 |
| linggao | :) | 15:00 |
| NobodyCam | ok bbiafm | 15:01 |
| jroll | agordeev: I hate circular imports :( | 15:03 |
| jroll | morning ironic! | 15:03 |
| agordeev | jroll: morning! | 15:08 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Replace sfdisk with parted https://review.openstack.org/84396 | 15:08 |
| jroll | :) | 15:08 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Replace sfdisk with parted https://review.openstack.org/84396 | 15:09 |
| * devananda tests the MP changes for parted | 15:12 | |
| devananda | https://review.openstack.org/#/c/86179/ | 15:13 |
| *** ilives has quit IRC | 15:13 | |
| devananda | *again | 15:13 |
| lucasagomes | heh will add tests to it | 15:14 |
| *** shausy has joined #openstack-ironic | 15:15 | |
| NobodyCam | mornig jroll :) | 15:15 |
| jroll | morning NobodyCam :) | 15:15 |
| devananda | lucasagomes: looks ilke the tripleo-ironic-seed test on that failed, but i think it was an error downloading a package | 15:16 |
| lucasagomes | devananda, yeah something was going on yesterday and it was failing pretty soon | 15:16 |
| lucasagomes | once I update the tests it will trigger the test again | 15:16 |
| lucasagomes | updated | 15:17 |
| lucasagomes | let's see | 15:17 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:23 |
| *** shausy has quit IRC | 15:24 | |
| *** jgrimm has joined #openstack-ironic | 15:24 | |
| *** overlayer has quit IRC | 15:30 | |
| devananda | hmmm, this is neat: http://activity.openstack.org/data/display/OPNSTK2/ironic | 15:31 |
| * NobodyCam clicks | 15:33 | |
| *** newell_ has joined #openstack-ironic | 15:33 | |
| NobodyCam | neet-o | 15:33 |
| NobodyCam | neat-0 even | 15:34 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Better check for available nodes https://review.openstack.org/84687 | 15:34 |
| lucasagomes | oh wow nice | 15:34 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Fix Nova rescheduling tear down problem https://review.openstack.org/84497 | 15:35 |
| dtantsur | lucasagomes, ^^^ | 15:35 |
| lucasagomes | dtantsur, :D will take a look, cherios | 15:35 |
| lucasagomes | cheerios* | 15:35 |
| dtantsur | Btw I do think tests there need some refactoring, e.g. to use @mock.patch or with mock.patch instead of long lines of .start() ... addClanup(mock.stop) | 15:36 |
| dtantsur | lucasagomes, also, did I use co-authored tag correctly? I'm confused it didn't turn into a neat link :) | 15:37 |
| lucasagomes | dtantsur, the tag seems grand | 15:38 |
| lucasagomes | Co-Authored-By: name <name@example.com> | 15:38 |
| dtantsur | cool! | 15:38 |
| lucasagomes | dtantsur, yeah the refactor is very needed, I think that we didn't know about the @mock decorator when i was writing the tests for the driver | 15:39 |
| lucasagomes | I remember that in the volume driver I added tests with the decorator | 15:39 |
| * lucasagomes looks the reviews in nova | 15:39 | |
| dtantsur | lucasagomes, and btw we already have https://review.openstack.org/#/c/73223/ and https://review.openstack.org/#/c/73256/ in progress | 15:39 |
| lucasagomes | dtantsur, yeah | 15:40 |
| *** mdickson has quit IRC | 15:40 | |
| russell_h | agordeev: is the ExecuteCommandMixin stuff you added just to work around the circular imports, or do you imagine wanting to add this functionality to other classes as well? | 15:40 |
| lucasagomes | heh yeah the nova driver doesn't use any @mock.patch | 15:40 |
| lucasagomes | https://review.openstack.org/#/c/51328/ :( | 15:40 |
| lucasagomes | but the volume does https://review.openstack.org/#/c/71026/15/nova/tests/virt/ironic/test_volume_driver.py | 15:41 |
| dtantsur | lucasagomes, would be nice to do before merging | 15:41 |
| lucasagomes | +1 | 15:41 |
| dtantsur | maybe I'll just create a dependent review request for this change tomorrow, and then we sync it into https://review.openstack.org/#/c/51328/ | 15:42 |
| *** dkehn_ is now known as dkehnx | 15:43 | |
| *** stevehuang has joined #openstack-ironic | 15:44 | |
| dtantsur | agordeev, I've added tests to https://review.openstack.org/#/c/84497/ as you requested :) | 15:45 |
| russell_h | agordeev: oh, you answered this in your commit message, nevermind | 15:47 |
| devananda | lucasagomes: not sure where it's coming from (possibly something I did wrong) but testing of the MP branch, i'm getting an instance with disk specs != the nova flavor | 15:50 |
| lucasagomes | devananda, scheduling problem? | 15:51 |
| lucasagomes | devananda, are the exact match filters enabled? | 15:51 |
| devananda | lucasagomes: there's only 1 node enrolledin ironic, and only one flavor in nova | 15:52 |
| *** mdickson has joined #openstack-ironic | 15:52 | |
| *** eghobo has joined #openstack-ironic | 15:52 | |
| devananda | lucasagomes: ah, this might be the "it creates 1GB of swap no matter what" thing? | 15:52 |
| devananda | lucasagomes: though i didnt know we had such a thing | 15:53 |
| lucasagomes | it creates 1mb swap no matter what | 15:53 |
| devananda | /dev/vda2 1305+ 1305- 1- 1024 82 Linux swap / Solaris | 15:53 |
| devananda | that's not 1mb | 15:53 |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Add FlowExtension https://review.openstack.org/85412 | 15:54 |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Resolve circular import, introduce ExecuteCommandMixin https://review.openstack.org/86341 | 15:54 |
| devananda | wait, yes it is | 15:54 |
| devananda | nvm :) | 15:54 |
| lucasagomes | heh | 15:54 |
| yuriyz | lucasagomes, please look at https://review.openstack.org/#/c/85086/ I want try +2 :) | 15:55 |
| lucasagomes | yeah check the size unit of that output | 15:55 |
| lucasagomes | yuriyz, hah ack will do | 15:55 |
| lucasagomes | yuriyz, cheers for the suggestion I will fix it quickly | 15:55 |
| devananda | lucasagomes: so sfdisk -u option isn't working in cirros ... | 15:55 |
| lucasagomes | devananda, I think cirros uses busybox | 15:56 |
| lucasagomes | so it's kinda limited types of commands | 15:56 |
| lucasagomes | not sure | 15:56 |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Add FlowExtension https://review.openstack.org/85412 | 15:56 |
| openstackgerrit | Aleksandr Gordeev proposed a change to openstack/ironic-python-agent: Resolve circular import, introduce ExecuteCommandMixin https://review.openstack.org/86341 | 15:56 |
| lucasagomes | btw, yuriyz rloo Haomeng, congrats! | 15:56 |
| lucasagomes | great to have you guys in the core team :) | 15:56 |
| yuriyz | thanks | 15:56 |
| rloo | thx lucasagomes! | 15:56 |
| devananda | sfdisk -u M -s /dev/vda2 | 15:56 |
| devananda | 1024 | 15:56 |
| devananda | that's really blocks, not MB | 15:57 |
| devananda | heh | 15:57 |
| devananda | ok, mystery solved | 15:57 |
| lucasagomes | :D | 15:57 |
| *** vkozhukalov has quit IRC | 15:58 | |
| NobodyCam | lol still no debug logging grrrrrrrr :-p | 15:59 |
| *** datajerk has quit IRC | 16:00 | |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: JsonPatch add operation on existing property https://review.openstack.org/85086 | 16:02 |
| lucasagomes | yuriyz, ^ here it go :D fixed and rebased | 16:02 |
| NobodyCam | huh... Can not change instance association while node 172834f2-95a6-4658-9f97-2884c3281ba6 is in power state power on. | 16:04 |
| devananda | lucasagomes: http://paste.openstack.org/show/75429/ | 16:06 |
| NobodyCam | ok now I really need to see what up. lucasagomes question this https://review.openstack.org/#/c/85529/5/ironic/conductor/manager.py should power off a node if its on when it added? | 16:06 |
| NobodyCam | devananda: also ^^^ | 16:06 |
| devananda | NobodyCam: hmm, interesting question. if we add a node, which is ON, and the conductor discovers that, which it should, we should still be able to manually power it off | 16:07 |
| devananda | NobodyCam: try ironic node-set-power-state UUID off | 16:07 |
| NobodyCam | this is in the gate tests | 16:07 |
| devananda | NobodyCam: i dont thin that should be automatic | 16:07 |
| * lucasagomes looking | 16:07 | |
| devananda | NobodyCam: yea, i get that tripleo leaves things running between tests - -to me, that's a bug in tripleo | 16:08 |
| devananda | NobodyCam: it's starting each round of tests in a dirty environment | 16:08 |
| NobodyCam | ya... ok I'll dig in when I get back | 16:08 |
| devananda | NobodyCam: and then not handling the situation it created. that's not a bug in ironic - -it's a deficiency in the test tooling that can't handle the situation it is creating | 16:08 |
| devananda | lifeless: ^ thoughts? | 16:08 |
| lucasagomes | hmm the seed vm depploys the undercloud and then tear it down | 16:09 |
| lucasagomes | I mean at least it should turn it off | 16:09 |
| lucasagomes | after the tests | 16:09 |
| *** datajerk has joined #openstack-ironic | 16:10 | |
| lucasagomes | unless we have a periodic task in Ironic that would turn the node off if it's not deployed or not in maintenance mode | 16:10 |
| devananda | lucasagomes: we have sync power state | 16:11 |
| devananda | but the issue is that on the next run of seed, the node is still powered on but has no prior state in the DB | 16:11 |
| devananda | so when it is first covered by the sync power state periodic task, its state is recorded in the db for the first time | 16:11 |
| lucasagomes | devananda, thats ur cirros image? | 16:12 |
| devananda | that patch is proposing that any node without a prior known state should be forced OFF, which I disagree with | 16:12 |
| devananda | lucasagomes: yea | 16:12 |
| devananda | lucasagomes: that paste is weird, no? | 16:12 |
| lucasagomes | devananda, yeah I think it's a tripleo problem as well, it's not turning the nodes off after the tests | 16:12 |
| lucasagomes | as I think it should | 16:12 |
| devananda | lucasagomes: i deployed cirros to a VM with the default flavor, then deleted the instance, changed the flavor, and deployed a new instance | 16:12 |
| devananda | lucasagomes: to see what parted would do | 16:12 |
| devananda | old flavor: root:10, swap:null, eph:null | 16:13 |
| lucasagomes | devananda, right, yeah looks a bit misaligned | 16:13 |
| devananda | new flavor: root:5gb, swap:2mb, eph:1gb | 16:13 |
| lucasagomes | devananda, I can't see the ephemeral there tho | 16:13 |
| devananda | right | 16:14 |
| lucasagomes | ohhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh I see | 16:14 |
| lucasagomes | the driver is not passing the ephemeral to ironic | 16:14 |
| lucasagomes | just like it wasn't passing the swap | 16:14 |
| * lucasagomes checks the maps | 16:14 | |
| devananda | ah | 16:14 |
| lucasagomes | https://github.com/openstack/ironic/blob/master/ironic/nova/virt/ironic/ironic_driver_fields.py | 16:15 |
| *** lsmola has quit IRC | 16:15 | |
| lucasagomes | no ephemeral | 16:15 |
| devananda | lucasagomes: yep. no ephemeral in driver_info | 16:15 |
| lucasagomes | devananda, you can set it directly in ironic to test | 16:15 |
| lucasagomes | pxe_ephemeral_gb and pxe_ephemeral_format | 16:15 |
| lucasagomes | pxe_ephemeral_gb=1 pxe_ephemeral_format=ext4 | 16:16 |
| lucasagomes | for e.g | 16:16 |
| *** martyntaylor has quit IRC | 16:16 | |
| lucasagomes | if u want to preserve it accross deployments u can do a pxe_preserve_ephemeral=true | 16:16 |
| lucasagomes | ok lemme fix it in the driver and test passing ephemeral | 16:16 |
| lucasagomes | I always test it but my scripts are adding it to ironic directly | 16:16 |
| lucasagomes | :( | 16:17 |
| lifeless | devananda: ? | 16:17 |
| devananda | lifeless: hi! was pinging you re: tripleo undercloud failure due to nodes being left powered on between test runs | 16:17 |
| devananda | lifeless: NobodyCam posted a potential fix that would cause ironic to force newly enrolled nodes OFF, rather than discovering their actual state, which I don't think we should do | 16:18 |
| lifeless | devananda: we power them off at the strt of the run | 16:19 |
| devananda | lucasagomes: seems like we're really pushing this feature far too late, yanno? | 16:19 |
| lifeless | devananda: but I like nobodycams patch | 16:19 |
| lucasagomes | devananda, :( | 16:19 |
| lucasagomes | devananda, yeah | 16:19 |
| lifeless | devananda: I was given a rack to bring up in useast; all the machines were on. | 16:20 |
| lifeless | devananda: I had to write a script to power them off | 16:20 |
| lifeless | devananda: btw powering them off is consistent with the power discussion we had a week or two back which was: | 16:20 |
| lifeless | - if there is an instance, nova determines desired state | 16:20 |
| lifeless | - if the node is in maintenance, admin determines desired state | 16:20 |
| *** datajerk has quit IRC | 16:21 | |
| lifeless | - if ironic wants the node on (e.g. to inventory) then its on | 16:21 |
| lifeless | - otherwise it is off | 16:21 |
| lifeless | devananda: unless you're proposing create-node should create it in mainenance by default | 16:21 |
| lucasagomes | oh no... lifeless I think it's fine to the auto-discovery to create it in maintenance mode, but it shouldn't be the default behavior for node-create | 16:22 |
| lifeless | lucasagomes: I don't either :) | 16:22 |
| lifeless | NobodyCam: see tripleo-incubator/scripts/boot-seed-vm line 146 | 16:23 |
| *** lazy_prince has joined #openstack-ironic | 16:25 | |
| devananda | lifeless: hm. right. so i'm thinking of the situation where an admin enrolls a bunch of nodes by eg. digestign a factory manifest, and the nodes suddenly power off | 16:25 |
| devananda | lifeless: without any log message or notice | 16:26 |
| *** lazy_prince is now known as killer_prince | 16:26 | |
| lifeless | devananda: and the admin goes 'fuck yeah, automation' | 16:26 |
| devananda | lifeless: heh | 16:26 |
| jroll | heh | 16:27 |
| jroll | and then /b 12 | 16:27 |
| jroll | oops >.> | 16:28 |
| lucasagomes | devananda, force_power_state_during_sync could be disabled if the admins want | 16:29 |
| lucasagomes | so it won't try to power off/on the node | 16:29 |
| lucasagomes | will update the db with the current state instead | 16:29 |
| devananda | lifeless: following this a step further, ironic should then refuse to turn on a node that is not either assoiated to an instance or in maintenance mode? | 16:31 |
| devananda | *do you think that | 16:31 |
| lifeless | devananda: or being turned on by ironic itself | 16:31 |
| lifeless | devananda: e.g. to inventory/firmware/whatever | 16:31 |
| devananda | right -- for internal reasons, eg. discovery | 16:31 |
| lifeless | devananda: and yes, I d othink that | 16:31 |
| devananda | eg, right now, a user can simply say "ironic node-set-power-state UUID on" | 16:31 |
| devananda | k | 16:31 |
| lifeless | devananda: I hadn't realised I thought that until you pointed it out :) | 16:32 |
| lifeless | but you can't turn on a libvirt domain that doesn't exist :) | 16:32 |
| lifeless | devananda: you *could* make node-set-power-state UUID on set maintenance mode on automatically | 16:32 |
| devananda | it's the conclusion of this approach. ironic pwns your hardware and you can't turn it on unless $things | 16:32 |
| russell_h | agordeev: could you pass the extension manager into the constructor? | 16:34 |
| *** datajerk has joined #openstack-ironic | 16:34 | |
| * russell_h is on his phone on a train, will be available for real in 20m | 16:35 | |
| *** killer_prince has quit IRC | 16:37 | |
| *** shausy has joined #openstack-ironic | 16:39 | |
| *** martyntaylor has joined #openstack-ironic | 16:40 | |
| JayF | What about for people who want nodes power state to remain on in order to facilitate quick deployment? | 16:40 |
| JayF | For example, we wanted to have the agent idle waiting for deploy requests so it didn't take two full power cycles to deploy a workload to a piece of metal | 16:40 |
| JayF | It seems like an always-off-by-default approach would prevent that? | 16:40 |
| *** vkozhukalov has joined #openstack-ironic | 16:41 | |
| jroll | JayF: I still (think) I agree that when a node is first registered with ironic, it should be powered off | 16:42 |
| *** ndipanov has quit IRC | 16:42 | |
| jroll | JayF: and I'm looking into working around the other bits like "node needs to be off to kick a deploy" | 16:43 |
| JayF | Cool, that obliterates my concern | 16:43 |
| jroll | :) | 16:43 |
| *** overlayer has joined #openstack-ironic | 16:48 | |
| *** harlowja_away is now known as harlowja | 16:49 | |
| *** overlayer has quit IRC | 16:50 | |
| *** yuriyz has quit IRC | 16:52 | |
| devananda | lucasagomes: even after delete & recreate with the same (second) flavor, it's still got CHS mismatch | 16:53 |
| lucasagomes | devananda, yeah it will create the partitions on the same offset.. hmm wondering if it's the optimal option that is causing it | 16:54 |
| lucasagomes | parted -a optimal | 16:54 |
| lucasagomes | devananda, what's the output of the sfdisk -d /dev/vda? | 16:57 |
| lucasagomes | would be nice to see the parted /dev/vda print as well but I think there's no parted on cirros | 16:59 |
| devananda | lucasagomes: http://paste.openstack.org/show/75433/ | 16:59 |
| devananda | yea, no parted | 16:59 |
| *** Alexei_987 has quit IRC | 17:02 | |
| *** lazy_prince has joined #openstack-ironic | 17:03 | |
| russell_h | agordeev: so IRC on phones sucks. Not sure if what I said came through, could we instantiate the appropriate extension manager and pass it in to the constructor? | 17:03 |
| russell_h | instantiate it prior to calling the constructor that is | 17:03 |
| *** lazy_prince is now known as killer_prince | 17:04 | |
| *** epim has joined #openstack-ironic | 17:06 | |
| lucasagomes | devananda, the fdisk -l seems pretty clear... idk if that alignment is harmful or not | 17:07 |
| lucasagomes | but maybe we should use -a cylinder instead of -a optimal | 17:07 |
| lucasagomes | to round it to cylinders | 17:07 |
| devananda | lemme patch that and try | 17:07 |
| lucasagomes | devananda, ack | 17:07 |
| lucasagomes | thanks | 17:07 |
| * devananda gets food and moves outside | 17:09 | |
| *** derekh has quit IRC | 17:10 | |
| *** jistr has quit IRC | 17:13 | |
| *** athomas has quit IRC | 17:16 | |
| *** coolsvap is now known as coolsvap|afk | 17:18 | |
| devananda | lucasagomes: didn't fix it. here's the cmd run: parted -a cylinder -s /dev/disk/by-path/ip-10.0.0.8:3260-iscsi-iqn-33cf7940- | 17:20 |
| devananda | 75f5-4dde-b940-17b7d58b32bf-lun-1 -- mklabel msdos unit MiB mkpart primary 1 5121 mkpart primary linux-swap 5121 5123 | 17:21 |
| *** pradipta_away is now known as pradipta | 17:21 | |
| devananda | though the result is slightly different | 17:22 |
| *** dwalleck has joined #openstack-ironic | 17:23 | |
| lucasagomes | devananda, right I got dig more into it | 17:23 |
| devananda | i'm not sure what problems this may cause | 17:24 |
| lucasagomes | yeah, that's my concern as well | 17:24 |
| lucasagomes | there's no overlap on the partitions which is good | 17:24 |
| lucasagomes | missaligment can affect performance | 17:24 |
| *** athomas has joined #openstack-ironic | 17:27 | |
| *** dwalleck has quit IRC | 17:27 | |
| *** overlayer has joined #openstack-ironic | 17:30 | |
| *** eguz has joined #openstack-ironic | 17:32 | |
| devananda | comstud: hi! are you going to follow up on rloo's suggestion on https://review.openstack.org/#/c/84822/1/ironic/api/controllers/v1/node.py ? | 17:34 |
| devananda | comstud: i'm fine merging this, just dont want the suggestion to get lost | 17:35 |
| *** epim has quit IRC | 17:36 | |
| *** eghobo has quit IRC | 17:36 | |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add lookup method to vendor mapping https://review.openstack.org/86402 | 17:36 |
| *** martyntaylor has left #openstack-ironic | 17:38 | |
| *** shausy has quit IRC | 17:40 | |
| *** shausy has joined #openstack-ironic | 17:40 | |
| openstackgerrit | A change was merged to openstack/ironic: JsonPatch add operation on existing property https://review.openstack.org/85086 | 17:45 |
| *** overlayer has quit IRC | 17:47 | |
| Shrews | rloo: ping | 17:48 |
| rloo | hi Shrews | 17:48 |
| Shrews | rloo: hi! quick question on your review 85107 | 17:49 |
| Shrews | rloo: there are several instances where you use 'ironic' rather than 'Ironic' when referring to the service (e.g.: 'ironic endpoint'). is that on purpose? | 17:49 |
| rloo | Shrews. you caught me. Well, not really on purpose. I can't remember when to use which. | 17:50 |
| devananda | comstud: looks like 84573 needs a rebase | 17:50 |
| rloo | Shrews. there is docn somewhere I believe, about when to capitalize. | 17:50 |
| rloo | Shrews. do you know? | 17:50 |
| Shrews | rloo: My personal opinion is that we should capitalize always when referring to the service itself. | 17:51 |
| devananda | comstud: actually, it's a real pep8 error. and i suspect it'll need a rebase too | 17:51 |
| rloo | Shrews. ok. let me see if I can find the docn on it. | 17:51 |
| lucasagomes | right I'm done for the day have a good night everyone | 17:51 |
| lucasagomes | devananda, I will investigate more tomorrow | 17:51 |
| devananda | lucasagomes: fwiw, the CHS offset issue isn't present when using sfdisk | 17:52 |
| lucasagomes | devananda, but I couldn't find anything really relevant to say it's harmful | 17:52 |
| lucasagomes | devananda, yeah tried that as well | 17:52 |
| lucasagomes | but it fails if you try to partition with the same inputs you had there | 17:52 |
| devananda | lucasagomes: yea. swap = 2mb kills it | 17:53 |
| devananda | heh | 17:53 |
| rloo | Shrews: https://wiki.openstack.org/wiki/Documentation/Conventions#Service_and_project_names | 17:53 |
| devananda | lucasagomes: seems like it fails with swap > 1024 sectors, not 1024 MB | 17:53 |
| lucasagomes | devananda, http://paste.openstack.org/show/75437/ | 17:53 |
| lucasagomes | yeah | 17:53 |
| lucasagomes | so sfdisk is doing something wrong I just can't figure out why | 17:53 |
| rloo | Shrews: My eyes glaze over reading "Generally the capitalization of the project team names like swift is lowercase. When used as a command-line-client name, lowercase is best. When used for the team or project name, typically it's not necessary to capitalize. " | 17:54 |
| lucasagomes | devananda, good reading http://www.ibm.com/developerworks/linux/library/l-4kb-sector-disks/ | 17:54 |
| Shrews | rloo: interesting! thanks for that. apparently my preference contradicts the standard :) | 17:55 |
| lucasagomes | devananda, so, idk the misalignment that sfdisk is showing might impact in performance but I think it's not a problem | 17:55 |
| devananda | lucasagomes: yes, a lot of the perf tuning I did while at percona was based on disk sector size and offset alignment | 17:55 |
| rloo | Shrews: I don't actually know if those conventions apply to developer docn ;) | 17:55 |
| lucasagomes | devananda, yeah | 17:55 |
| lucasagomes | devananda, also, partioning with >=8GB seems to not cause that misaligment showed by sfdisk | 17:56 |
| lucasagomes | sudo parted -a cylinder -s /dev/vdb -- mklabel msdos unit MiB mkpart primary 1 $((5*1024+1)) mkpart primary linux-swap $((5*1024+1)) $((5*1024+3) | 17:56 |
| lucasagomes | 5 is the number of GB | 17:56 |
| openstackgerrit | Russell Haering proposed a change to openstack/ironic: Version agent lookup payloads https://review.openstack.org/85228 | 17:56 |
| JoshNang | hey guys, how do i run the tests for the nova driver? | 17:56 |
| *** epim has joined #openstack-ironic | 17:57 | |
| *** epim has quit IRC | 17:57 | |
| lucasagomes | devananda, I will investigate it more tomorrow | 17:58 |
| *** lucasagomes is now known as lucas-dinner | 17:58 | |
| lucas-dinner | night devananda NobodyCam rloo JoshNang Shrews etc... | 17:58 |
| JoshNang | g'night lucas-dinner! | 17:58 |
| devananda | lucasagomes: g'night! i'll keep looking into it as i also dig through the review backlog :) | 17:59 |
| rloo | g'night lucas-dinner! | 17:59 |
| lucas-dinner | devananda, ack, thanks! lemme know by gtalk or email if you find something so I can continue in the morning | 17:59 |
| Shrews | JoshNang: they should run automatically under the py26 and py27 tests now | 17:59 |
| Shrews | lucas-dinner: night! | 18:00 |
| *** datajerk has quit IRC | 18:00 | |
| JoshNang | Shrews: was that a recent (last couple days) change? i have a test that's first line is "raise Exception()" but passes all tests | 18:00 |
| Shrews | JoshNang: yes, very recent | 18:00 |
| JoshNang | kk thanks! | 18:00 |
| *** epim has joined #openstack-ironic | 18:02 | |
| rloo | devananda: fyi, this *should* fix the ironicclient docn: https://review.openstack.org/#/c/86390/ | 18:03 |
| *** datajerk has joined #openstack-ironic | 18:03 | |
| devananda | rloo: nice cleanup of the layout.yaml section | 18:11 |
| devananda | rloo: while you're touching that, what do you think about template "name: check-requirements" ? | 18:12 |
| rloo | devananda: thx. btw, do you know what the check-requirements does? I noticed that ironic had it, but was too lazy to look into it. | 18:12 |
| rloo | devananda: i was just going to ask you about that. | 18:12 |
| devananda | heh | 18:12 |
| rloo | devananda. so I should, or you are asking me if I know whether we should or not? | 18:13 |
| devananda | rloo: i believe it should be there | 18:13 |
| devananda | rloo: it enforces common versions of things in rquirements.txt | 18:14 |
| devananda | and test-requirements.txt | 18:14 |
| rloo | devananda: ok then, I'll add it in to the same review. | 18:14 |
| devananda | thanks! | 18:14 |
| devananda | JoshNang: you probably need to invoke the test suite with "tox -epy27", from outside the venv | 18:15 |
| devananda | JoshNang: or if you want to run just the nova unit tests, do what tox is doing directly... | 18:15 |
| devananda | something like bash -c "TESTS_DIR=./ironic/nova/tests/ python setup.py testr --slowest --testr-args='{posargs}'" | 18:15 |
| JoshNang | devananda: thanks! not sure why it wasn't work yesterday. heh i never run it inside a venv...it usually decided to rebuild the venv first then | 18:16 |
| devananda | (but just running tox is better) | 18:16 |
| devananda | yea | 18:17 |
| *** Mikhail_D_ltp has joined #openstack-ironic | 18:19 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding a reference driver for the agent https://review.openstack.org/84795 | 18:22 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add lookup method to agent vendor mappings https://review.openstack.org/86402 | 18:31 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add lookup method to agent vendor mappings https://review.openstack.org/86402 | 18:33 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Use configurable driver name in lookup URL. https://review.openstack.org/86201 | 18:34 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Use configurable driver name in lookup URL. https://review.openstack.org/86201 | 18:35 |
| rloo | In Monday's meeting, someone was asking about devstack/ironic/fedora? Here's the etherpad I mentioned: https://etherpad.openstack.org/p/jjWcLDThTK | 18:37 |
| rloo | but I don't recall who wrote it and I don't know how up-to-date it is. | 18:38 |
| *** max_lobur1 has quit IRC | 18:42 | |
| openstackgerrit | A change was merged to openstack/ironic: Fix self.fields on API Node object https://review.openstack.org/84822 | 18:44 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic: Add version 1 of agent payload https://review.openstack.org/86173 | 18:48 |
| openstackgerrit | A change was merged to openstack/ironic: Sync base object code with Nova's https://review.openstack.org/84859 | 18:49 |
| *** datajerk has quit IRC | 18:50 | |
| *** epim has quit IRC | 18:55 | |
| *** epim has joined #openstack-ironic | 18:59 | |
| *** pradipta is now known as pradipta_away | 19:05 | |
| *** epim has quit IRC | 19:05 | |
| *** shausy has quit IRC | 19:10 | |
| NobodyCam | me is almost back | 19:12 |
| openstackgerrit | Jarrod Johnson proposed a change to stackforge/pyghmi: Allow pyghmi to send break over IPMI SOL https://review.openstack.org/86425 | 19:13 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Use configurable driver name in lookup URL https://review.openstack.org/86201 | 19:17 |
| NobodyCam | lifeless: line 146 is "fi" ???? | 19:18 |
| openstackgerrit | Jarrod Johnson proposed a change to stackforge/pyghmi: Allow pyghmi to send break over IPMI SOL https://review.openstack.org/86425 | 19:20 |
| *** derekh has joined #openstack-ironic | 19:29 | |
| *** epim has joined #openstack-ironic | 19:29 | |
| openstackgerrit | A change was merged to stackforge/pyghmi: Allow pyghmi to send break over IPMI SOL https://review.openstack.org/86425 | 19:31 |
| *** openstackgerrit has quit IRC | 19:34 | |
| *** datajerk has joined #openstack-ironic | 19:36 | |
| *** Mikhail_D_ltp has left #openstack-ironic | 19:36 | |
| russell_h | agordeev: I think we can probably drop the command_name thing | 19:36 |
| russell_h | agordeev: at one point we were constructing the command result object inside the actual handler, so we needed the name | 19:37 |
| russell_h | I'm not sure if we still do that anywhere | 19:37 |
| devananda | Shrews: this - return self._multi_getattr(client, method)(*args) - do you think it may obfuscate dependency issues if/when the ironicclient library changes? | 19:37 |
| russell_h | agordeev: yeah, pretty sure that can go, the async_command decorator has obviated the need for it | 19:38 |
| devananda | Shrews: eg, it'll raise AttributeError, but that's not handled anywhere afaict | 19:38 |
| russell_h | agordeev: wait, wrong guy, sorry | 19:39 |
| Shrews | devananda: if it raises that exception, then either it's a programmer error (that should never be merged) or the api has changed in an incompatible way | 19:40 |
| Shrews | devananda: i'm not crazy about the whole using-strings-for-the-call thing either, but it's what nova does. i can't come up with a better way | 19:40 |
| devananda | Shrews: heh | 19:41 |
| russell_h | vkozhukalov: ^ see what I was saying to agordeev, that was meant for you | 19:41 |
| comstud | devananda: ok. ya, i have some follow up stuff, I need to push new versions for them all | 19:41 |
| lifeless | NobodyCam: yes, but look at the code block :) | 19:42 |
| *** openstackgerrit has joined #openstack-ironic | 19:42 | |
| devananda | Shrews: so "api has changed in an incompatible way" -- why we'll eventually need upgrade testing, too | 19:43 |
| NobodyCam | lifeless: yes. we looking at that | 19:43 |
| lifeless | NobodyCam: it turns off the instances, no ? OH - I see, we don't for ironic. right, because Ironic is meant to ;) | 19:43 |
| devananda | lifeless: re the sfdisk/parted discussion, if you have a moment, does http://paste.openstack.org/show/75433/ raise any red flags for you? | 19:44 |
| devananda | lifeless: i'm curious what your take on the CHS mismatch is | 19:45 |
| vkozhukalov | russell_h: no, I don't think so, it is certainly not for me, I just suggested how we can deal with uninitialized ext_mgr | 19:46 |
| NobodyCam | devananda: I see your comments on the fix undercloud patch.. if I read correctly you are okay with the power off if it is logged? (i was planning to remove all the debug stuff from the patch) | 19:46 |
| *** max_lobur has joined #openstack-ironic | 19:46 | |
| russell_h | vkozhukalov: I must be totally losing it :) | 19:47 |
| russell_h | vkozhukalov: wait, I'm still sane. Maybe. I'm talking about this: https://review.openstack.org/#/c/85412/7/ironic_python_agent/flow.py,unified | 19:47 |
| vkozhukalov | russell_h: ahh, yes. I just saw that my Pycharm fades this argument 'command_name' in all methods where it appears | 19:49 |
| devananda | NobodyCam: yes. it's a reasonable option to enable, but needs to be clearly doc'd (both in the option string and in our eventual deployer docs) and logged | 19:49 |
| vkozhukalov | russell_h: so, it seems not to be needed at all | 19:50 |
| devananda | NobodyCam: it'd be helpful to also stick a DocImpact flag in the commit message | 19:50 |
| devananda | NobodyCam: so we can start grepping our git logs for things that need to be called out at release doc time | 19:50 |
| NobodyCam | devananda: ack let me rework that patch | 19:51 |
| russell_h | vkozhukalov: right, before we had this decorator we had to construct some object inside those methods which needed the command name | 19:51 |
| russell_h | vkozhukalov: we kept it in the signature for consistency, but IMO we can kill it now | 19:51 |
| russell_h | vkozhukalov: I'll reply to the change | 19:51 |
| vkozhukalov | russell_h: thx | 19:52 |
| lifeless | devananda: CHS? | 19:53 |
| lifeless | devananda: you know thats entirely legacy these days right? | 19:54 |
| lifeless | devananda: oh, I see the errors | 19:54 |
| lifeless | devananda: yes, badish partition table | 19:54 |
| devananda | lifeless: clearly, it's a VM i'm working with. but that :) | 19:54 |
| devananda | so this is easy to reproduce with lucas' parted patch by simply deploying again to a node, with a flavor that has a different root_gb than the previos instance | 19:55 |
| lifeless | hahaha :(. parted. | 19:56 |
| devananda | right | 19:56 |
| devananda | lifeless: in tjhe tripleoci cloud, are you creating any instances with nova-bm that have non-zero swap partitions? | 20:00 |
| devananda | lifeless: cause i dont see any diference in the sfdisk cmds for novabm and ironic, which suggests that https://bugs.launchpad.net/ironic/+bug/1297925 should apply to nova too | 20:00 |
| lifeless | devananda: it does, and we're not. | 20:02 |
| devananda | :( | 20:03 |
| lifeless | frankly I'd just list it as a known defect and focus on other stuff :) | 20:03 |
| devananda | i had hoped this patch fixed it, which would have been nice ... | 20:03 |
| devananda | but now i suspect this issue with parted applies to all of lucas' parted work on master, too, not just the backport | 20:04 |
| devananda | so, yea | 20:04 |
| lifeless | I missed 'partition twice' as a test fo rthe test matric. | 20:05 |
| lifeless | dammit | 20:05 |
| *** eguz has quit IRC | 20:07 | |
| *** eghobo has joined #openstack-ironic | 20:07 | |
| devananda | ? | 20:08 |
| lifeless | devananda: when I reviewed the parted patch | 20:08 |
| devananda | ah | 20:08 |
| devananda | heh | 20:08 |
| lifeless | I described a test matrix we needed | 20:08 |
| lifeless | I didn't think to say 'oh and make sure it edits correctly' | 20:08 |
| lifeless | because we're not *meant* to be editing | 20:08 |
| devananda | + partition again with a different partition table | 20:08 |
| devananda | well, it doesn't need to "edit" -- it should erase & create new | 20:09 |
| devananda | but it should do that without errors | 20:09 |
| lifeless | right | 20:09 |
| devananda | which it doesn't :( | 20:09 |
| openstackgerrit | Ruby Loo proposed a change to openstack/python-ironicclient: Documentation for contributors https://review.openstack.org/85096 | 20:11 |
| russell_h | jroll: https://review.openstack.org/#/c/85228/ | 20:11 |
| * jroll looks | 20:14 | |
| jroll | lgtm | 20:15 |
| *** openstackstatus has quit IRC | 20:15 | |
| *** openstackstatus has joined #openstack-ironic | 20:16 | |
| *** athomas has quit IRC | 20:23 | |
| *** romcheg has quit IRC | 20:31 | |
| *** epim has quit IRC | 20:32 | |
| *** harlowja is now known as harlowja_away | 20:36 | |
| devananda | anyone else care to give https://review.openstack.org/#/c/84497/ a review? | 20:36 |
| devananda | backport potential, fixes an issue where you cant delete a nova instance that previously failed to deploy | 20:37 |
| jroll | oh hey, I have that issue | 20:39 |
| * jroll takes a look | 20:39 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Adding swift temp url support https://review.openstack.org/81391 | 20:44 |
| *** max_lobur1 has joined #openstack-ironic | 20:46 | |
| *** max_lobur has quit IRC | 20:47 | |
| devananda | lol. just spotted this: man sfdisk ... BUGS: there are too many options. | 20:48 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 20:49 |
| *** romcheg has joined #openstack-ironic | 20:49 | |
| *** linggao has quit IRC | 20:50 | |
| *** jistr has joined #openstack-ironic | 20:50 | |
| NobodyCam | devananda: how the log message look in manager.. also changes help string and added DocImpact | 20:53 |
| *** jdob has quit IRC | 20:54 | |
| *** harlowja_away is now known as harlowja | 20:55 | |
| *** mrda_away is now known as mrda | 20:55 | |
| NobodyCam | morning mrda :-p | 20:55 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added lvm partitioner https://review.openstack.org/85672 | 20:56 |
| mrda | Morning NobodyCam | 20:57 |
| mrda | and everyone else :) | 20:57 |
| devananda | NobodyCam: help string is better. but the log is not doing what I suggested // think it should | 20:58 |
| *** BadCub01 has joined #openstack-ironic | 20:58 | |
| devananda | NobodyCam: do you see any reason not to log a different message based on whether force-node-power-state is set? | 20:58 |
| NobodyCam | I wrote that like tree times and it just didn't look right to me | 20:59 |
| devananda | vkozhukalov: lvm in the agent? jroll ?? | 20:59 |
| NobodyCam | three even | 21:00 |
| devananda | maybe my surprise is a surprise to you, but managing LVM seems to be stepping outside of "provisioning" | 21:00 |
| devananda | and into the responsibilities of the tenant | 21:00 |
| jroll | devananda: hey | 21:01 |
| *** jistr has quit IRC | 21:01 | |
| *** epim has joined #openstack-ironic | 21:01 | |
| NobodyCam | I can make the log different based on force_power_state_during_sync | 21:01 |
| jroll | devananda: does nova allow for things like this? (thinking no) | 21:01 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk partitioner https://review.openstack.org/86163 | 21:01 |
| jroll | devananda: I can see it *almost* fitting but I'm not sure | 21:01 |
| devananda | jroll: i believe nova uses lvm to provide functionality outside of an instance (eg, instance snapshotting) | 21:04 |
| devananda | jroll: but it's not mucking around inside the instance with lvm | 21:04 |
| vkozhukalov | devananda: why not? what if I want image to be copied on lv partition not on a plain one? | 21:04 |
| jroll | devananda: ok, hmm | 21:05 |
| *** derekh has quit IRC | 21:05 | |
| devananda | side note, are you guys using """ for file headers instead of # ??? | 21:09 |
| devananda | eg, https://review.openstack.org/#/c/86163/3/ironic_python_agent/disk_partitioner.py | 21:09 |
| devananda | wont that show up in the doc builds, which you *dont* want? | 21:10 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 21:10 |
| jroll | so, we originally used """ | 21:10 |
| NobodyCam | devananda: ^^?? | 21:10 |
| jroll | we've been (slowly) moving away from that | 21:10 |
| jroll | I thought we got rid of them but /me will doublecheck | 21:10 |
| JoshNang | there's probably still a few | 21:10 |
| devananda | right, so vkozhukalov's patches are adding more | 21:11 |
| devananda | i just commented | 21:11 |
| jroll | oh wow, there's still plenty in the agent | 21:11 |
| devananda | heh | 21:11 |
| jroll | I'll take care of that | 21:11 |
| devananda | so yea, ya'll should move away from that, and add some sphinx docs too :) | 21:11 |
| devananda | (eventually) | 21:12 |
| jroll | indeed | 21:12 |
| *** eghobo has quit IRC | 21:12 | |
| NobodyCam | doh | 21:12 |
| jbjohnso_ | woo I can send break like crazy from CLI now | 21:13 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 21:13 |
| jbjohnso_ | one day I'll get up the bother to bestow that gift unto the javascript client somehow | 21:13 |
| NobodyCam | jbjohnso_: sounds like a dos attack waiting to happen | 21:13 |
| NobodyCam | lol | 21:13 |
| jbjohnso_ | NobodyCam, I mean 'send break' | 21:13 |
| vkozhukalov | devananda: so, what about lvm? I think it is critical for us to have lvm and md functionality in the agent | 21:14 |
| jbjohnso_ | NobodyCam, aka 'hurray, a way to request a stack trace from moderately hung kernel' | 21:14 |
| devananda | vkozhukalov: how will the partition type (lv, msdos, gpt, etc) be determined? | 21:14 |
| *** openstackstatus has quit IRC | 21:14 | |
| jbjohnso_ | devananda, fyi, there is something of a scary bomb in the TianoCore EFI implementation with respect to partition tables | 21:15 |
| jbjohnso_ | devananda, if something that is GPT today gets touched by something that isn't GPT but yet doesn't *completely* eradicate all bits of the GPT MBR... | 21:15 |
| jbjohnso_ | devananda, TianoCore helpfully decides 'oh, you must not have meant that, here, let me overwrite your superblock with a GPT' | 21:16 |
| russell_h | devananda: vkozhukalov: IMO we need to standardize a deploy api ASAP | 21:16 |
| jbjohnso_ | hilarity ensues | 21:16 |
| russell_h | today the API is something like "lol ask the driver" | 21:16 |
| vkozhukalov | devananda: it should be completely data driven (json) | 21:16 |
| russell_h | which pushes these sorts of questions down to the implementation | 21:17 |
| devananda | NobodyCam: better. one nit, commented | 21:17 |
| NobodyCam | :) | 21:17 |
| jroll | devananda: no copyright on empty files, right? | 21:17 |
| devananda | jroll: right | 21:18 |
| jroll | thanks | 21:18 |
| vkozhukalov | russell_h: yes, it should be like "lol ask the driver", i believe everyone wants to have their own deployment scenario, that is why agordeev is working on flow driver | 21:18 |
| jbjohnso_ | empty files are precious intellectual property | 21:18 |
| devananda | russell_h: demonstrate the API, at the very least | 21:18 |
| devananda | russell_h: doubtless it'll take at least a little time for folks to hammer on it and figure out how things fit into it | 21:19 |
| NobodyCam | ahh good catch fixing now | 21:19 |
| devananda | russell_h: as there are multiple use cases for the agent | 21:19 |
| openstackgerrit | Vladimir Kozhukalov proposed a change to openstack/ironic-python-agent: Added disk partitioner https://review.openstack.org/86163 | 21:19 |
| *** openstackstatus has joined #openstack-ironic | 21:22 | |
| russell_h | vkozhukalov: I've been led to believe that the broader openstack community isn't going to be very receptive to having drivers expose different APIs | 21:22 |
| devananda | jroll: now taht instance_info is available in ironic, plans to migrate nova.virt.ironic and ironic.drivers.modules.pxe to use it? | 21:23 |
| *** jbjohnso_ has quit IRC | 21:23 | |
| devananda | jroll: also, https://review.openstack.org/#/c/85131/2 looks like it probably depends on other work. is this dependency not strict and therefor not represented in gerrit? | 21:23 |
| russell_h | vkozhukalov: for example, Nova doesn't have different APIs for different virt drivers | 21:24 |
| russell_h | vkozhukalov: it abstracts the problem such that many drivers can implement the same API | 21:24 |
| *** romcheg has quit IRC | 21:25 | |
| devananda | russell_h: exactly | 21:25 |
| *** epim has quit IRC | 21:25 | |
| jroll | devananda: yes, I can migrate those | 21:25 |
| jroll | devananda: on that review, things won't break if the agent driver doesn't exist | 21:25 |
| jroll | devananda: if that's what you mean by "not strict" | 21:26 |
| devananda | whether partitioning is done by software in the agent (lvm, parted, etc) or in the PXE driver (over an iSCSI conenction) or out-of-band (eg, via some iLO magic) or what ever... the API neesd to be the same | 21:26 |
| *** notq has quit IRC | 21:26 | |
| devananda | jroll: exactly. ok. but for things that don't make sense in isolation (eg, there are a bunch of related but not linearly-dependent patches) could you | 21:27 |
| JoshNang | anyone run into the doc gate check quitting with "IOError: [Errno 28] No space left on device"? http://logs.openstack.org/91/81391/9/check/gate-ironic-docs/e4b16d4/console.html | 21:27 |
| vkozhukalov | russell_h: I did not mean we can have different APIs for drivers, well defined API does not contradict with sending many different parameters to many different drivers | 21:27 |
| devananda | jroll: tag the later ones as WIP, with a comment pointing to the earlier ones? | 21:27 |
| devananda | jroll: that'll help me when i'm reviewing to find the head of the changesets | 21:27 |
| devananda | jroll: also using a consistent topic is really great | 21:27 |
| devananda | eg, i can search for related patches like this: https://review.openstack.org/#/q/status:open+project:openstack/ironic+branch:master+topic:add-agent-virt-fields,n,z | 21:28 |
| vkozhukalov | russell_h: I mean POST + json, that is exactly what we disscussed in sunnyvale | 21:28 |
| devananda | if they all have the same topic, it's super helpful | 21:28 |
| devananda | vkozhukalov: sure it does. if the client has to send different parameters based on the back-end impilementation, the abstraction is lost | 21:28 |
| devananda | vkozhukalov: imagine if you had to issue a completely different set of aprameters to "nova boot" if the back end was libvirt or xen.... | 21:29 |
| vkozhukalov | russell_h: good example is patching driver_info in ironic, i can set almost anything there, how does it break the abstract of well defined API? | 21:30 |
| devananda | JoshNang: nice... just "recheck no bug" on that patachset | 21:30 |
| devananda | vkozhukalov: it does. heh. the whole driver_info usage today is a broken abstraction :) | 21:30 |
| devananda | vkozhukalov: this is why we're movign that to instance_info | 21:30 |
| JoshNang | :) thanks | 21:31 |
| devananda | vkozhukalov: all fields in instance_info will be standardized, and used to determine things like partition layout and size | 21:31 |
| devananda | vkozhukalov: whereas the driver_info should be set only once, by the operator, so that ironic is able to control the hardware | 21:31 |
| openstackgerrit | Jim Rollenhagen proposed a change to openstack/ironic-python-agent: Use # instead of """ for copyright blocks https://review.openstack.org/86455 | 21:32 |
| jroll | devananda: sure, do you want me to fix the existing reviews as well? | 21:34 |
| vkozhukalov | devananda: ok, I see your point. I'll think of it and will suggest something. but what I really need to know right now is - Does having lvm contradicts with ironic at all and it's use cases? I'm asking because it is very very critical for comprehensive deployments | 21:34 |
| vkozhukalov | devananda: if I can not have lvm in agent, so I doubt if it is useful for me at all. | 21:35 |
| devananda | jroll: if you don't mind, i think that'll help make it a bit clearer to me (and other reviewers) | 21:36 |
| *** epim has joined #openstack-ironic | 21:36 | |
| jroll | sure | 21:37 |
| devananda | jroll: since you have a lot of related patche up, and it's nto clear from gerrit how they're related | 21:37 |
| jroll | yep | 21:38 |
| jroll | np | 21:38 |
| devananda | vkozhukalov: how do you plan on / need to use lvm? | 21:38 |
| *** datajerk has quit IRC | 21:38 | |
| devananda | vkozhukalov: particularly, after ironic has deployed an instance to a node, why does it matter if it was with lvm or gpt style partition? | 21:39 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 21:40 |
| vkozhukalov | devananda: for example, I want to be able to have huge partition which consists of many hard drives. I use lvm for that. | 21:40 |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Handling validation in conductor consistently https://review.openstack.org/84374 | 21:40 |
| *** max_lobur1 has quit IRC | 21:42 | |
| devananda | vkozhukalov: some of my concern is that lvm is generally used from *within* a host on an ongoing basis to manage volume groups | 21:42 |
| devananda | vkozhukalov: it's significantly more user-facing than, say, md or MegaRaid, and its use often extends long after provisioning the system is complete | 21:43 |
| devananda | whereas building the RAID and creating a partition table are clearly prerequisites to putting the OS on it, and part of "provisioning" | 21:44 |
| devananda | manaing LV groups is not | 21:44 |
| devananda | *managing | 21:44 |
| * NobodyCam reads MegaRaid and runs to the corner and cowers | 21:45 | |
| devananda | vkozhukalov: so how do we separate "build the RAID" from "manage logical volume groups" ? | 21:45 |
| devananda | haha | 21:45 |
| *** datajerk has joined #openstack-ironic | 21:45 | |
| NobodyCam | :) | 21:46 |
| devananda | sorry, NobodyCam. I had allll kinds of fun doing performance tuning of those cards later ;) | 21:46 |
| vkozhukalov | devananda: ok, imaging i have 1G 2 hard drives, my OS just won't fit such a small disk, but having lvm allows me to solve this issue | 21:46 |
| NobodyCam | :) i can bet | 21:46 |
| NobodyCam | thou I wouldn't use the excat word "fun" | 21:47 |
| devananda | vkozhukalov: sure. why not use mdadm? | 21:47 |
| russell_h | devananda: can't images with lvm partitions on them? | 21:47 |
| russell_h | er, can't you have | 21:47 |
| devananda | russell_h: at the moment, ironic only spports partition images | 21:48 |
| devananda | russell_h: when we haev support for whole disk images, sure, it's possible | 21:48 |
| devananda | russell_h: actually THAT would be ok, from ironic's POV, since it is just putting the requested image on disk | 21:49 |
| russell_h | right | 21:49 |
| NobodyCam | oh poop.. forgot to regen the conf sample | 21:49 |
| devananda | my concern is conflating the scope of ironic into user space by saying "we'll create LVM stuff for you. oh, but you're on your own after that, since ironic won't manage the PV/LV groups, snapshots, etc" | 21:50 |
| *** datajerk has quit IRC | 21:50 | |
| devananda | then we get two more problems | 21:50 |
| vkozhukalov | devananda: I can't see the difference between lvm and md according to the case when user want to shoot his own leg | 21:50 |
| devananda | user changes something in the PV, then redeploys their node, but the partition table is different | 21:50 |
| vkozhukalov | devananda: both those tools are of the same kind | 21:50 |
| devananda | or, user asks why ironic created an LVM group but then refuses to manage it | 21:51 |
| devananda | vkozhukalov: no, they're not. LVM does *a lot* more than mdadm | 21:51 |
| devananda | unless mdadm gained a whole lot of features while i wasn't looking | 21:51 |
| russell_h | devananda: the same is sort of true of mdadm though, right? As far as managing it post-deployment | 21:51 |
| devananda | actually | 21:51 |
| russell_h | like Ironic is never going to rebuild your raid | 21:51 |
| devananda | right | 21:52 |
| devananda | they're both softwre layer raids | 21:52 |
| devananda | we can write some aspects of it to the partition table, but if the OS doesn't honor that, what's the point? | 21:52 |
| vkozhukalov | devananda: ok, should we make filesystem? they also do a lot more than just storing inodes | 21:52 |
| russell_h | I'm not super opinionated on this and I'm all about reducing scope, but I can sort of see an argument for LVM | 21:52 |
| vkozhukalov | devananda: ok, bad example, filesystem is a part of image | 21:53 |
| russell_h | vkozhukalov: beyond dropping an image onto the node, IMO Ironic shouldn't touch filesystems | 21:53 |
| jroll | JoshNang: can you mark this WIP with a pointer to the dependent patches in a comment? https://review.openstack.org/#/c/84795/ | 21:55 |
| *** epim has quit IRC | 21:56 | |
| JoshNang | jroll: yup. didn't realize it lost WIP status with new patch sets | 21:56 |
| *** datajerk has joined #openstack-ironic | 21:57 | |
| jroll | devananda: once that one is marked, all of the agent-related patches should have dependencies pointed at in the "depends on" section or a WIP comment | 21:58 |
| devananda | so w.r.t. mdadm, i was thinking initially of "fake raid" systems where there is still hardware assistance in mapping the drives together, but no raid acceleration | 21:58 |
| devananda | i dont see how ironic can deploy any purely software raid solution, without being very tightly coupled to the image that is being deployed | 21:59 |
| devananda | russell_h: re: "ironic is never going torebuild your raid" -- well. if it's a hardware raid controller, ironic _should_ be able to rebuild it, change raid level, etc | 22:02 |
| JoshNang | but it wouldn't be able to rebuild a software raid | 22:03 |
| russell_h | devananda: sure, I meant mdadm | 22:03 |
| JoshNang | heh | 22:03 |
| russell_h | devananda: but IMO that is OK | 22:03 |
| russell_h | devananda: like we're dropping an image onto a disk, but that doesn't imply we're going to upgrade packages for you | 22:04 |
| JayF | We create a starting point for a person/automation to put a workload on the system, and until the tenant gives it back up, it's their responsibility, right? | 22:04 |
| JayF | if that starting point involves LVM, or MD RAID, how does that imply anything different? | 22:05 |
| russell_h | IMO, unless explicitly stated otherwise, it should be implied that between the time a deploy completes and the time you destroy it, Ironic isn't going to manage your server at all | 22:05 |
| JayF | with maybe the exception that you have to be using an image that supports it | 22:05 |
| devananda | russell_h: right | 22:06 |
| devananda | JayF: how is that determined? | 22:06 |
| JayF | devananda: how is what? if the image supports it? | 22:06 |
| devananda | JayF: how does ironic know that it should use mdadm, or lvm, or not, for a given image? | 22:06 |
| devananda | JayF: by comparison, if the hardware has a hw raid card that IPA can configure, and the user has requested (eg, via cinder, or an as-yet-unknown API) that a RAID5 be built, then ironic should do that before putting the image down | 22:07 |
| devananda | and that RAID is idnependent of the image | 22:07 |
| devananda | but if I take taht same image, and put it into a partition created on top of LVM, what happens? | 22:08 |
| devananda | (that's sort of not a rhetorical question) | 22:08 |
| JayF | I don't know the direct answer to your question, but there are already implicit requirements in images -- for instance, you can't deploy a RHEL5 image to a machine that has hardware too new for RHEL5 to work -- which is a problem unique to Ironic vs other nova drivers, correct? | 22:08 |
| devananda | JayF: driver issues are, IMO, fairly well understood -- and addressed by disk-image-builder | 22:09 |
| JayF | Isn't this similarly a driver issue though? In the case of mdadm/lvm, if it's configured properly, the only difference is having the relevant packages installed and baked into the initramfs (which most distros do for you) | 22:10 |
| JayF | I'm just trying to figure out where the line is, and how those problems are already being solved. | 22:11 |
| devananda | will the system's firstboot / init scripts detect the LV group and configure itself appropriately? | 22:12 |
| JayF | I can't speak for every distribution/image, but generally speaking, yes | 22:12 |
| devananda | maybe -- i haven't tested it, but that's only half of my concern | 22:12 |
| devananda | the other half is scope | 22:12 |
| JayF | same for MD, if the raid has the superblocks configured properly | 22:12 |
| JayF | I don't personally have a desire to do this, just want to understand why not | 22:13 |
| devananda | LVM, in my experience, is typically managed -- and frequently changed -- post install of the OS | 22:13 |
| devananda | let's imagine a user does this | 22:13 |
| JayF | Heh. Mine has been actually almost exactly the opposite. It's configured at install time and not touched anymore. | 22:13 |
| devananda | - request node with 1TB disk and a 100GB root partition | 22:14 |
| devananda | - ironic delivers taht | 22:14 |
| devananda | actually, nvm | 22:14 |
| devananda | that line breaks down regardless of whether its LVM or something else | 22:15 |
| JayF | I especially don't see the difference between hardware raid vs software raid+lvm | 22:15 |
| devananda | JayF: how about - i can put a windows image on a hardware raid, but not on a software raid+lvm | 22:16 |
| JayF | That's a good reason; but what about GPT vs MBR at that point? | 22:16 |
| JayF | hmm. I guess that isn't a valid argument *now* because windows supports both | 22:17 |
| JayF | So using 'can this work for linux and windows' as a guideline really knocks a few things out of scope, and helps me understand where the line is. | 22:17 |
| JayF | Although it's sad to think we'll lose the ability to do some things because Windows can't do it, it's a reasonable place to draw the line. I get it, thanks. | 22:18 |
| russell_h | even if we accept that as where we draw the line right now, we should try to design an API that can be extended to move the line later | 22:19 |
| devananda | JayF: or how about "do my preprovisioning steps depend on whether it's a RHEL, Debian, or XXX" image? | 22:19 |
| vkozhukalov | does nova supports lxc as a hypervisor? does windows supports lxc as a hypervisor? | 22:19 |
| russell_h | like, I don't think I've ever used hardware RAID in my life | 22:19 |
| JayF | russell_h: you've used servers that had hardware raid, for sure. I managed them :P | 22:19 |
| russell_h | it'd be a shame for Ironic not to support software raid | 22:19 |
| devananda | russell_h: that's exactly the point -- you used it transparently. | 22:20 |
| devananda | vkozhukalov: your question doesn't make sense. are you comparing nova to windows? | 22:20 |
| russell_h | JayF: sure, and I regretted everything about that experience ;) | 22:20 |
| russell_h | JayF: except the part where you managed them, of course | 22:20 |
| JayF | haha | 22:20 |
| vkozhukalov | devananda: no, I just want to say that it is not a valid argument if windows does not support lvm | 22:20 |
| russell_h | I would have killed to just have generic hardware and software raid | 22:21 |
| russell_h | and not have these clowns trying to pitch me different models of crap at every turn | 22:21 |
| devananda | vkozhukalov: a better argument would be, "if an image can run on nova+kvm, should we assume it will also run on nova+xen" | 22:22 |
| vkozhukalov | devananda: ok, why not nova+kvm and nova+lxc ?? the same question about comparing nova + kvm and nova + BM. | 22:23 |
| vkozhukalov | devananda: bm is not a hypervisor, it is a real world | 22:23 |
| JayF | If you got a Windows image to run in lxc, I'd be really impressed. :P | 22:23 |
| devananda | vkozhukalov: um, I think the same image DOES work for nova+kvm and nova+xen and nova+ironic :) | 22:24 |
| devananda | vkozhukalov: ironic functions very similarly to a hypervisor | 22:25 |
| devananda | vkozhukalov: and it should, IMO, not reach into the user / tenant environment. I know you guys wanted to have an agent always running inside the tenant environment | 22:26 |
| devananda | vkozhukalov: adding LVM support to Ironic and then using an in-guest agent to manage that is really neat stuff | 22:26 |
| devananda | and I think would be super useful | 22:26 |
| devananda | but | 22:26 |
| devananda | i think that breaks out of the scope for ironic | 22:27 |
| devananda | on the other hand, providing software raid AND hardware raid pre-provisioning, underneath a single abstraction | 22:27 |
| devananda | is really an awesome thing | 22:27 |
| devananda | and maybe within the scope of Ironic ... but it's a slippery slope ... | 22:27 |
| JayF | I think a good middle ground would be yes, but probably not now | 22:28 |
| devananda | how much does this make ironic depend on the *content* of the image? | 22:28 |
| vkozhukalov | devananda: ok, I remember you wrote in one of blueprints where we discussed discovering that we can not discovery hardware via IPMI or something like and we need to do that from inside ramdisk OS. do you see the difference? i always can discover Virtual machine configuration outside of it, but it does not work for BM. | 22:28 |
| JayF | devananda: I'd argue that the hardware /already/ depends on the content of the image. | 22:28 |
| devananda | JayF: yes, but *ironic* does not | 22:28 |
| devananda | JayF: the operator knows what hardware they have and provides a reasonable image to ironic that supports (eg has the right drivers) for that hardware | 22:28 |
| devananda | JayF: ironic doesn't inject new drivers into the image | 22:29 |
| devananda | (yea, someone suggested that a while back) | 22:29 |
| NobodyCam | ieek | 22:29 |
| *** radsy has joined #openstack-ironic | 22:29 | |
| NobodyCam | we should never crack open a image | 22:29 |
| *** openstack has joined #openstack-ironic | 22:34 | |
| * NobodyCam goes back to reading the scroll back | 22:34 | |
| devananda | JayF: it's possible, with extra image metadata | 22:34 |
| russell_h | devananda: it seems like almost regardless, if you want to configure RAID, you'd need to either allow the user to pass in some specification or have it hardcoded on a per-hardware basis | 22:34 |
| JayF | devananda: I don't know -- but I do think it's valuable to reconsider if we should declare s/w raid and lvm out of scope. Although 'later' would be a good answer for me as well :) | 22:35 |
| *** matty_dubs is now known as matty_dubs|lunch | 22:35 | |
| JayF | russell_h: that's basically what I was thinking, yeah | 22:35 |
| *** matty_dubs|lunch is now known as matty_dubs|gone | 22:35 | |
| *** datajerk has quit IRC | 22:35 | |
| russell_h | devananda: in either case, as you say, the operator will need to supply an appropriate image | 22:35 |
| NobodyCam | JayF: russell_h: devananda: we could have unconf talk about it? | 22:36 |
| jroll | reviews already exist in gerrit for this stuff, I'd prefer not to let them hang for over a month | 22:36 |
| NobodyCam | ack | 22:37 |
| devananda | one thing that'll go a long way for me - -demonstrate that the same image works with and without an LVM raid | 22:38 |
| *** jgrimm has quit IRC | 22:38 | |
| russell_h | perhaps vkozhukalov could put up a blueprint describing how this would look to a user | 22:39 |
| vkozhukalov | yes I'll do that | 22:40 |
| russell_h | ie, does this require some sort of specification at deploy time, or does the operator need to somehow hardcode RAID configurations | 22:40 |
| russell_h | and if you specify it at deploy time, what does it look like | 22:40 |
| russell_h | including both the JSON and considerations about interactions with images | 22:40 |
| devananda | that ^ goes back to "we need to start spec'ing a common API for all the agent stuffs" | 22:40 |
| russell_h | yeah :) | 22:40 |
| russell_h | I just like looking at JSON | 22:41 |
| devananda | and that I believe that aPI needs to abstract the particular implementation details | 22:41 |
| russell_h | right | 22:41 |
| devananda | eg, whether it's hardware or software raid, the user interaction, the nova-> ironic messages, all need to be the same | 22:41 |
| devananda | as another tangent | 22:42 |
| devananda | all the performance tuning for AWS users that I did | 22:42 |
| devananda | they all used RAID on top of EBS | 22:43 |
| russell_h | yeah | 22:43 |
| devananda | normal workflow: request instance with small root partition. attach larger storage system(s). build LVM on top of that. | 22:43 |
| devananda | and folks already want to attach cinder volumes to Ironic nodes | 22:44 |
| devananda | and they will then LVM on top of those | 22:44 |
| russell_h | right, IMO Ironic should focus on the root partition, like I don't want to be raiding together data disks and shit | 22:44 |
| devananda | sure, but | 22:44 |
| devananda | now teh user has two different tools that create(d) their LVM groups | 22:44 |
| devananda | thus my sense that it's a layer violation | 22:45 |
| JayF | An LVM group is an LVM group though? | 22:45 |
| JayF | What does it matter who or what created them for the root disk vs the data disk[s]? | 22:45 |
| devananda | how would it look if you requested an AWS instance and it came with an LVM group already? | 22:45 |
| devananda | i'm sort of comparing apples to oranges ... i know ... | 22:46 |
| JayF | It'd look especially bad because I'd be a Racker using AWS ;) | 22:46 |
| devananda | JayF: hah! | 22:46 |
| JayF | but I'd enjoy the ability to call the API for the Rackspace Cloud, request LVM root, and get it | 22:46 |
| JayF | and then would have my OS-level config management setup LVM on the data drive | 22:46 |
| pquerna | just to throw out how i tend to view this as a layering violation | 22:47 |
| pquerna | like, what happens when my OS is windows | 22:47 |
| pquerna | is the api going to support its software raiding ntfs stuff? | 22:47 |
| devananda | :) | 22:47 |
| JayF | pquerna: but what if a user wants to automatically deploy images on baremetal on software raid? They just can't? Or is there some other layer that would handle this? | 22:48 |
| pquerna | i'm not sure. | 22:48 |
| pquerna | i just view doing a 'ton' of things in this space as.... scary. | 22:48 |
| pquerna | there probally is a minimal ammount that is needed | 22:49 |
| pquerna | and maybe some way to make plugins | 22:49 |
| pquerna | but the 'baseline' should stive to be simple | 22:49 |
| pquerna | and maybe that means, root device is just the glance image DD'ed over | 22:49 |
| pquerna | then once the OS boots, it can mess with stuff | 22:49 |
| pquerna | now, that only gets messy if you want a root raid5/10/0 | 22:50 |
| pquerna | or ZFS or something of that kind | 22:50 |
| pquerna | and LVM | 22:50 |
| pquerna | right, so its a mess | 22:50 |
| comstud | mm zfs, let's do it. | 22:50 |
| devananda | mmm, btrfs | 22:50 |
| pquerna | right, and that. | 22:50 |
| pquerna | and windows software raid | 22:50 |
| pquerna | man, thats a big ass matrix already | 22:50 |
| pquerna | i'm liking DD to root device only by default :x | 22:51 |
| devananda | right. so. when the line was "ironic does HARDWARE things", it was much clearer. why aren't we sticking to that again? | 22:51 |
| comstud | i think you build the images how you want them | 22:51 |
| comstud | have an LVM image if you really want | 22:51 |
| comstud | and a non-LVM image | 22:51 |
| devananda | cause supporting hp, ibm, dell, amd, etc... isn't nearly a complex enough matrix already ;) | 22:51 |
| pquerna | so, the 'line' for hardware in this case, is a way for a plugin, to execute megaraid-blahblah | 22:52 |
| pquerna | as part of its build for a flavor | 22:52 |
| devananda | the "prepare the hardware to be suitable for this flavor" line | 22:52 |
| pquerna | but all the formatting, expanding of partitions, making /var 10gb instead of 80gb, is up to the image | 22:52 |
| devananda | which could include HW raid, change BIOS setting, flash firmware... | 22:53 |
| pquerna | its a bummer about hw raid. | 22:53 |
| *** datajerk has joined #openstack-ironic | 22:54 | |
| pquerna | i guess, the one case i'd like to understand is this: would you ever flash firmware differently | 22:54 |
| pquerna | based on the Image ID? | 22:54 |
| devananda | pquerna: the question that vkozhukalov and jayf have raised is, should that plugin to support megaraid also support a plugin for lvm | 22:54 |
| pquerna | or is hw raid just this one exception | 22:54 |
| devananda | pquerna: based on image? no. based on flavor? yes. | 22:54 |
| pquerna | right, based on flavor i'm all in on. | 22:55 |
| pquerna | but based on the image || create params, we are saying, yeah, maybe run megaraid in two possible configs raid10 root or raid5 root, as an example | 22:55 |
| pquerna | are those options of raid5 || raid10 part of the properties of the flavor? | 22:55 |
| devananda | they could be | 22:56 |
| pquerna | so, maybe its really, flavors can have a couple 'enum' choices | 22:56 |
| pquerna | like, flavor Blah has disk_layout: raid10,raid5 | 22:56 |
| devananda | "swift-flavor" might be JBOD. "trove-flavor" might be RAID10 | 22:56 |
| devananda | right | 22:56 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add configdrive generation to Nova driver https://review.openstack.org/86192 | 22:57 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic: Add agent fields to virt driver https://review.openstack.org/85131 | 22:57 |
| devananda | well, i was expecting more that a flavor expresses a single set of configs | 22:57 |
| pquerna | hmm | 22:57 |
| pquerna | i'd expect it would be more like: big-disk-flavor-type: disk_layout: JBOD|RAID10 .... create(image_id=$swift_image, disk_layout=JBOD) | 22:58 |
| pquerna | then the agent just gets a create like normal, if a plugin is loaded that understands disk_layout for megaraid, it hooks in, executes megaraid | 22:59 |
| devananda | pquerna: the current expectation is that flavor == hardware profile | 22:59 |
| pquerna | right | 22:59 |
| pquerna | ah, i see | 22:59 |
| pquerna | hrm. | 22:59 |
| pquerna | so a flavor would imply _all_ hw settings | 23:00 |
| devananda | right | 23:00 |
| pquerna | are you suggesting that you could have multiple flavor..... types on a node? | 23:00 |
| pquerna | and there is a 'change flavor type' command to the agent? | 23:00 |
| pquerna | (just thinking in irc here) | 23:00 |
| devananda | oh | 23:00 |
| devananda | not exactly | 23:00 |
| devananda | i'm thinking that multiple flavors could be matched to the same node | 23:00 |
| devananda | and when the user requests flavor X and image Y, that any available node which _can_ match flavor X is chosen, and udpated if necessary | 23:01 |
| devananda | and then image Y written to it | 23:01 |
| devananda | whereas a node N could be matched by flavor A and B, and the difference might be A {'disk layout': JBOD, 'vt': off} and B {'disk layout': RAID10, 'vt': on} | 23:02 |
| pquerna | i mean, seems good. not sure about how you want to exress that node N can be flavor A or B. | 23:02 |
| devananda | but otherwise bothflavors might have the same cpu/ram/disk properties | 23:03 |
| devananda | extra specs | 23:03 |
| pquerna | feels scary | 23:03 |
| devananda | currently scheduler matches on both capabilities (cpu/ram/disk) and extra specs | 23:03 |
| pquerna | like its just random/key/value stuff | 23:03 |
| devananda | yep | 23:03 |
| pquerna | have i ever mentioned how much i had json fields. | 23:03 |
| pquerna | lemme find a gif to express it | 23:03 |
| russell_h | no | 23:03 |
| pquerna | s/had/hate/ | 23:04 |
| devananda | heh | 23:04 |
| russell_h | let me find an email where you told me to make something a JSON field | 23:04 |
| * russell_h probably has hundreds | 23:04 | |
| devananda | so it's a stunningly beautiful afternoon outside and i haven't stepped away from this laptop for the last 9 hours | 23:04 |
| pquerna | http://i.imgur.com/vy3IXHz.gif | 23:04 |
| pquerna | i think thats about it for jsonfields | 23:04 |
| devananda | thanks for the very good discussion, everyone | 23:12 |
| devananda | let's keep it going (later). it's probably worth a move to the ML at this point | 23:12 |
| openstackgerrit | A change was merged to openstack/ironic: Fix seamicro power.validate() method definition https://review.openstack.org/85675 | 23:15 |
| *** datajerk has quit IRC | 23:22 | |
| NobodyCam | woo hoo /me finally get tests running on os x | 23:22 |
| devananda | russell_h: https://review.openstack.org/#/c/84142/ and https://review.openstack.org/#/c/84814/2 | 23:23 |
| devananda | russell_h: looks like yuriy's patch directly overlaps yours. can yhou comment? | 23:23 |
| russell_h | devananda: yep, thanks | 23:24 |
| russell_h | devananda: I'm alright with either merge order, although I believe that Yuriy's patch is a strict superset of mine | 23:26 |
| russell_h | so if we think mine is good to go we should just pull the trigger | 23:26 |
| devananda | superset? you mean subset? | 23:26 |
| russell_h | er | 23:26 |
| russell_h | yes | 23:26 |
| russell_h | if we think mine needs more love, lets merge his, its a trivial rebase | 23:26 |
| * russell_h comments | 23:26 | |
| devananda | ooh | 23:27 |
| devananda | one thing your patch is missing | 23:27 |
| devananda | test_responsible_for_node | 23:27 |
| russell_h | true that | 23:28 |
| russell_h | ok, lets merge Yuriy's then, I'll rebase on top | 23:28 |
| devananda | k | 23:28 |
| openstackgerrit | Chris Krelle proposed a change to openstack/ironic: Fix for tripleO undercloud gate tests https://review.openstack.org/85529 | 23:29 |
| *** lucas-dinner has quit IRC | 23:29 | |
| JoshNang | so i broke a couple tempest test trying to fix the db details leak bug https://review.openstack.org/#/c/73121/. where do i go to fix those? i'm not too familiar with how temptest works. | 23:29 |
| adam_g | Shrews, you still able to reproduce that test suite deadlock on demand? | 23:30 |
| devananda | ok, /me really goes for a walk | 23:30 |
| devananda | also have dinner plans, so i'm probably done for the day ... we'll see... might hop on later | 23:30 |
| JoshNang | have a good night! | 23:30 |
| adam_g | devananda, cheers | 23:31 |
| NobodyCam | have a good night devananda | 23:31 |
| *** datajerk has joined #openstack-ironic | 23:38 | |
| * NobodyCam wanders away too.. will check on test in a while | 23:40 | |
| *** vkozhukalov has quit IRC | 23:41 | |
| openstackgerrit | Adam Gandelman proposed a change to openstack/ironic: Cleanup running conductor services in tests https://review.openstack.org/86473 | 23:54 |
| adam_g | Shrews, ^ give that a shot if so.. | 23:54 |
| adam_g | thats an unimpressive diff considering how long it took to track down. :) | 23:55 |
| openstackgerrit | Ruby Loo proposed a change to openstack/python-ironicclient: Adds documentation for ironicclient API https://review.openstack.org/85107 | 23:57 |
| JoshNang | adam_g: is that for the mem leak bug? | 23:57 |
| adam_g | JoshNang, yeah | 23:57 |
| JoshNang | if that fixes it, you'll be my hero | 23:57 |
| adam_g | JoshNang, give it a shot | 23:57 |
| adam_g | fixes it for me | 23:57 |
| JoshNang | kk. its been very sporadic for me | 23:58 |
| JoshNang | i'll test it out | 23:58 |
| adam_g | JoshNang, i was able to reproduce 100% by running discover manually in the venv | 23:58 |
| adam_g | discover -t ./ ./ironic/tests/ | 23:58 |
| jroll | wow | 23:59 |
| jroll | nice find, adam_g | 23:59 |
| *** killer_prince has quit IRC | 23:59 | |
| adam_g | JoshNang, but yeah--i could see how it wouldbe sporatic depending on how fast the tests run + when conductor services start up + when periodic tasks pop | 23:59 |
| adam_g | jroll, :) | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
