Monday, 2014-01-27

« Sunday, 2014-01-26 Index Tuesday, 2014-01-28 »
*** matsuhashi has joined #openstack-ironic00:09
Haomengmax_lobur: great work:)00:30
*** matsuhashi has quit IRC00:33
*** matsuhashi has joined #openstack-ironic00:48
*** nosnos has joined #openstack-ironic00:58
*** jcooley_ has joined #openstack-ironic01:12
*** anniec has joined #openstack-ironic01:24
*** anniec_ has joined #openstack-ironic01:25
*** anniec has quit IRC01:28
*** anniec_ is now known as anniec01:29
*** nosnos_ has joined #openstack-ironic01:53
*** nosnos has quit IRC01:54
openstackgerritA change was merged to openstack/ironic: Remove unused config option - pxe_deploy_timeout  https://review.openstack.org/6799002:26
*** jbjohnso has joined #openstack-ironic02:26
*** matsuhashi has quit IRC02:26
*** matsuhashi has joined #openstack-ironic02:32
*** matsuhashi has quit IRC03:18
*** anniec has quit IRC03:58
*** coolsvap has joined #openstack-ironic04:38
*** matsuhashi has joined #openstack-ironic04:40
*** jbjohnso has quit IRC04:41
*** jcooley_ has quit IRC04:50
*** jcooley_ has joined #openstack-ironic05:11
*** jcooley_ has quit IRC05:19
*** jcooley_ has joined #openstack-ironic05:20
*** jcooley_ has quit IRC05:24
*** jcooley_ has joined #openstack-ironic06:01
openstackgerritJenkins proposed a change to openstack/ironic: Imported Translations from Transifex  https://review.openstack.org/6802406:04
*** jcooley_ has quit IRC06:14
*** mdurnosvistov has joined #openstack-ironic07:17
*** lsmola_ has joined #openstack-ironic07:31
*** coolsvap has quit IRC07:39
*** coolsvap has joined #openstack-ironic08:21
*** mdurnosvistov has quit IRC08:29
*** yuriyz has joined #openstack-ironic08:33
*** matsuhashi has quit IRC08:36
*** matsuhashi has joined #openstack-ironic08:37
*** ifarkas has joined #openstack-ironic08:40
*** jistr has joined #openstack-ironic08:48
*** Alexei_987 has joined #openstack-ironic09:07
*** derekh has joined #openstack-ironic09:18
*** ndipanov has joined #openstack-ironic09:20
*** mdurnosvistov has joined #openstack-ironic09:34
*** athomas has joined #openstack-ironic09:43
*** mdurnosvistov has quit IRC09:45
*** mdurnosvistov has joined #openstack-ironic09:47
openstackgerritDmitry Shulyak proposed a change to openstack/ironic: alembic with initial migration and tests  https://review.openstack.org/6741509:50
*** tatyana has joined #openstack-ironic09:56
*** mdurnosvistov_ has joined #openstack-ironic10:01
*** mdurnosvistov has quit IRC10:03
*** martyntaylor has joined #openstack-ironic10:04
openstackgerritDmitry Shulyak proposed a change to openstack/ironic: alembic with initial migration and tests  https://review.openstack.org/6741510:07
*** athomas has quit IRC10:16
*** max_lobur_afk is now known as max_lobur10:18
*** athomas has joined #openstack-ironic10:23
*** romcheg has joined #openstack-ironic10:45
*** eglynn has joined #openstack-ironic10:55
*** coolsvap_away has joined #openstack-ironic11:05
*** coolsvap has quit IRC11:06
*** aignatov_ is now known as aignatov11:07
*** lucasagomes has joined #openstack-ironic11:18
lucasagomesHaomeng, morning/afternoon :)11:27
*** coolsvap_away is now known as coolsvap11:36
*** matsuhashi has quit IRC11:40
*** matsuhashi has joined #openstack-ironic11:45
*** matsuhashi has quit IRC12:12
*** matsuhashi has joined #openstack-ironic12:13
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Sync db.sqlalchemy, test and py3kcompat code from Oslo  https://review.openstack.org/6900412:26
*** coolsvap has quit IRC12:36
*** viktors has joined #openstack-ironic12:40
openstackgerritRoman Vyalov proposed a change to openstack/ironic: Add test spec for building ironic  https://review.openstack.org/6935212:52
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Add lockfile>=0.8 to requirements.txt  https://review.openstack.org/6935812:54
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Sync db.sqlalchemy, test and py3kcompat code from Oslo  https://review.openstack.org/6900413:00
viktorslucasagomes: hi!13:04
lucasagomesviktors, hey13:04
viktorslucasagomes: how are you?13:05
lucasagomesviktors, all is buddy :) urself?13:05
lucasagomesall is well*13:05
viktorslucasagomes: same things :)13:06
*** aignatov is now known as aignatov_13:06
*** max_lobur is now known as max_lobur_afk13:07
viktorslucasagomes: as for your fix with Oslo - probably you need fix tox.ini as well https://github.com/openstack/nova/blob/master/tox.ini#L1713:08
lucasagomesviktors, a-ha... cheers for that13:08
lucasagomesIt also failed because of a missing dependency on lockfile13:09
viktorslucasagomes: and can you please update code one more time? We just got https://review.openstack.org/#/c/59433/ patch merged, so we can move to alembic in ironic13:09
lucasagomesviktors, heh sure, just will wait the result of jenkins and then I update it again13:10
lucasagomesviktors, thanks :)13:10
viktorslucasagomes: ok, sure13:10
viktorslucasagomes: no problem :)13:10
lucasagomes:)13:10
*** vkozhukalov has joined #openstack-ironic13:20
*** eglynn has quit IRC13:23
lucasagomesviktors, oddly, when I run the update.py script from oslo to sync the new db.sqlachemy to Ironic it's not copying the migration_cli/ directory13:24
lucasagomesis that expected to be copied as well?13:25
Haomenglucasagomes: morning:)13:25
viktorslucasagomes: yes, let me check13:25
lucasagomesviktors, okie13:25
lucasagomesHaomeng, :) how r u?13:25
Haomenglucasagomes: fine, enjoy my vacation:)13:26
lucasagomesHaomeng, ah, nice!13:26
Haomenglucasagomes: :)13:26
lucasagomesHaomeng, can I annoy u a bit with some work related stuff?13:26
lucasagomesjust an updated about the progress on implemeting the IPMI-based monitoring driver/agent for Ceilometer13:27
Haomenglucasagomes: sure13:27
lucasagomesdo you have any news about it? the blueprint is assigned to you13:27
Haomenglelucasagomes: let me check13:27
lucasagomesHaomeng, cheers mate13:27
Haomenglucasagomes: is this one - https://blueprints.launchpad.net/ironic/+spec/send-data-to-ceilometer ?13:28
lucasagomesHaomeng, yes13:29
lucasagomesit's because a ceilometer dev came to talk to me today about it13:29
Haomenglucasagomes: now we have two solutions proposed, about the data model to integrate with ceilometer13:29
lucasagomesbut I had little info, so I told him I will investigate and poke few people13:29
lucasagomes(eglynn his name)13:30
viktorslucasagomes: yes you right. Looks like, that this can be bug in Oslo...13:30
viktorslucasagomes: Anyway, its works for me, after I added `module=db.sqlalchemy.migration_cli` line to openstack-common.conf13:30
Haomenglucasagomes: so should to be confirmed with ceilometer guys about our solutions, which one can be accepted by ceilometer team13:30
lucasagomesviktors, gotcha, will give it a try13:30
Haomenglucasagomes: I try to discuss with ceilometer guys previous days13:30
lucasagomesHaomeng, I see, but it still planned to be included in i3?13:31
Haomenglucasagomes: however not sure who is the ceilometer owner for this ipmi-data collection bp13:31
lucasagomesor you think it's going to take more time to integrate it ?13:31
Haomenglucasagomes: that is targed by Deva I think13:31
lucasagomes(i3 finishs at 6 of march I think13:31
lucasagomesHaomeng, gotcha13:31
Haomenglucasagomes: ok, let me try to commit first draft patch and let ceilometer team to review together13:32
lucasagomesHaomeng, that would be great, but no hurry, if u wanna do it after ur holidays feel free to do so13:32
Haomenglucasagomes: but not sure if we need more discusstion with ceilometer team officially13:32
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Removes use of timeutils.set_time_override  https://review.openstack.org/6743213:32
Haomenglucasagomes: i plan to do some code  in my vacation days if i am free:)13:33
lucasagomesHaomeng, I see :)13:33
lucasagomesHaomeng, next monday u'll be working already?13:33
lucasagomesI can try to talk to eglynn to come to our meeting and have a chat with us13:33
Haomenglucasagomes: but my concern here is, who is ceilomter ipmi data collection bp owner?13:33
lucasagomeshe's ceilometer core, so that would be good to iron out some of the problems13:33
lucasagomesHaomeng, no idea13:34
Haomenglucasagomes: great13:34
lucasagomeslemme check13:34
Haomenglucasagomes: now we proposed two solutions in  our ironic bp13:34
Haomenglucasagomes: just need to be confirmed with ceilometer team, what data format is accepted by them, and ceilometer have to do some work to handle the ipmi data13:34
Haomenglucasagomes: should I send mail to dev list?13:35
lucasagomesHaomeng, that would be useful13:35
lucasagomesI will follow up with him as well13:35
Haomenglucasagomes: maybe it is not required, I think we just involved ceilometer guys who focus on this bp13:35
Haomenglucasagomes: ipmi data is a lot of key-value pairs which retrieved by ipmi command13:36
lucasagomesI see13:36
* lucasagomes trying to find the ceilometer bp13:37
lucasagomeshttps://blueprints.launchpad.net/ceilometer/+spec/monitoring-physical-devices13:37
lucasagomesHaomeng, cheers for all the info, very useful13:38
*** aignatov_ is now known as aignatov13:38
Haomenglucasagomes: :)13:38
Haomenglucasagomes: that should be my responsibility:)13:39
Haomenglucasagomes: I will focus this bp from now13:39
lucasagomesHaomeng, ^^ no worries it's grand, you depend on the ceilometer guys work as well13:39
lucasagomesso u kinda blocked13:39
lucasagomesI will try to call their attention to that as well13:40
Haomenglucasagomes: I just have concern, why ceilometer have no such bp for ipmi?13:40
lucasagomesHaomeng, I've no idea, that monitoring-physical-devices mention that13:40
Haomenglucasagomes: and from our Ironic team, we need to co-work or discuss with the ceilometer ipmi bp owner together13:40
Haomenglucasagomes: np13:40
lucasagomes+113:41
Haomenglucasagomes: so for now, we need to find the guys which will be ceilometer ipmi-data-collection bp owner, but this not sure if this is monitoring-physical-devices bp13:41
Haomenglucasagomes: :)13:42
lucasagomesHaomeng, +1, yea I will check that out with eglynn and see if he has any inputs about it13:43
Haomenglucasagomes: thank you lucas13:44
Haomenglucasagomes: I will try to prepare the code from our ironic part13:44
lucasagomesHaomeng, np thank YOU :)13:44
Haomenglucasagomes: :)13:44
lucasagomesHaomeng, cheers, but no hurry, enjoy ur holiday as well13:44
Haomenglucasagomes: my body is in vacation, my heart is in ironic:)13:45
lucasagomes:D13:45
Haomenglucasagomes: :)13:45
Haomenglucasagomes: I have to save time to get more deep understanding about our project13:45
Haomenglucasagomes: :)13:46
lucasagomesHaomeng, :) you're doing well buddy and it takes time (we are under rapid development)13:47
HaomengHaomeng: now I think the gap is that ceilometer has its own data model, in ceilometer domain, the data model is - resource -> meter -> sample13:48
Haomenglucasagomes: so how to define our IPMI data model for ceilometer, that is point for both our ceilometer and ironic team13:48
Haomenglucasagomes: Lucas, I have another information - https://review.openstack.org/#/c/51828/13:50
Haomenglucasagomes: this is ceilometer  ipmi inspector , has been canceled by our ironic, so we have current bp owned by me, this is history13:51
* lucasagomes looks13:51
Haomenglucasagomes: this ipmi inspector ipmlemented the ipmi data colection for temperature, fan's speed, voltage three meters, that is mapping to our first solution13:52
Haomenglucasagomes: because ipmi is an open interface, so there are more data excluding these basic temperature, fan's speed, voltage information, so we have another common interface which is our solution 213:53
HaomengHaomeng: let me update our bp clearly to have a summary13:54
lucasagomesI see, that would be useful13:54
Haomenglucasagomes: ok13:55
*** eglynn has joined #openstack-ironic13:59
*** jbjohnso has joined #openstack-ironic14:05
*** romcheg1 has joined #openstack-ironic14:06
openstackgerritDmitry Shulyak proposed a change to openstack/ironic: alembic with initial migration and tests  https://review.openstack.org/6741514:07
*** eglynn is now known as eglynn-afk14:07
*** romcheg has quit IRC14:07
*** matty_dubs|gone is now known as matty_dubs14:08
*** romcheg has joined #openstack-ironic14:11
*** rustlebee is now known as russellb14:12
*** romcheg1 has quit IRC14:13
*** romcheg1 has joined #openstack-ironic14:25
*** romcheg has quit IRC14:27
*** max_lobur_afk is now known as max_lobur14:45
*** aignatov is now known as aignatov_14:52
*** eglynn-afk has quit IRC14:56
*** jdob has joined #openstack-ironic14:56
*** ndipanov has quit IRC15:08
*** rwsu has joined #openstack-ironic15:16
*** max_lobur_cellul has joined #openstack-ironic15:23
*** max_lobur_cellul has quit IRC15:24
Haomenglucasagomes: I have updated bp - https://blueprints.launchpad.net/ironic/+spec/send-data-to-ceilometer15:32
*** ndipanov has joined #openstack-ironic15:33
Haomenglucasagomes: can you help to get ceilometer guys to take a look, and we need ceilometer team input/comments15:33
*** aignatov_ is now known as aignatov15:33
lucasagomesHaomeng, cheers Haomeng, I talked to eglynn as well and he's going to join us on the next Ironic meeting (feb 3) where we can try to iron out the problems you mentioned15:33
lucasagomesHaomeng, will do as well :)15:34
lucasagomesthanks15:34
HaomengHaomeng: ok15:34
*** lucasagomes is now known as lucas-hungry15:35
Haomenglucasagomes: how do you think if we can discuss with eglynn before the meeting via irc?15:35
lucas-hungryHaomeng, he's usually online in this channel15:35
Haomenglucas-hungry: because our meeting is time limited15:35
lucas-hungryI think he's on PTO today (although he was answering the emails)15:36
lucas-hungrymaybe we should bring it up at the #openstack-ceilometer channel as well15:36
Haomenglucas-hungry: but maybe I can not attend these meetings during my holidays:)15:37
Haomenglucas-hungry: yes15:37
Haomenglucas-hungry: so I think if i can discuss with him via irc directly15:37
lucas-hungryHaomeng, I see, right15:37
lucas-hungryyea if you can talk to him directly that would be grand as well15:37
Haomenglucas-hungry: the meeting time is 1 hour limitted, no time to discuss in deep15:38
lucas-hungryindeed15:38
lucas-hungrywe always have some follow ups in this channel after the meeting15:38
Haomenglucas-hungry: but I am not sure how to find him, in ceilometer irc?15:38
lucas-hungryHaomeng, yes15:38
lucas-hungryor this channel15:38
lucas-hungrybut he's not online right now15:38
Haomenglucas-hungry: :)15:38
lucas-hungryi was talking to him via email15:38
lucas-hungryI think he's on PTO (holiday) today15:39
Haomenglucas-hungry: ok, let me sent mail to him and cc you and deva:)15:39
Haomenglucas-hungry: ok15:39
Haomenglucas-hungry: I dont think we should send this mail to dev list15:39
Haomenglucas-hungry: this should be our ceilometer and ironic team issue only:)15:39
*** coolsvap has joined #openstack-ironic15:40
lucas-hungryHaomeng, I see, well the devlist is fine for that as well15:40
lucas-hungryjust put the tags correctly15:40
Haomenglucas-hungry: and I have no idea what is more Effective:)15:40
lucas-hungry[Ironic][Ceilometer]15:40
Haomenglucas-hungry: how can i get his mail?15:40
lucas-hungrypm15:40
Haomengllucas-hungry: let me try, if you dont know:)15:40
HaomengHaomeng: is this nick name -  eglynn,  right?15:42
lucas-hungryyup15:42
lucas-hungrysent his email to u in a pm15:42
Haomenglucas-hungry: thk15:42
lucas-hungrynp15:42
Haomenglucas-hungry: ok, i will go to sleep:)15:42
Haomenglucas-hungry: and will not attend today's meeting:)15:42
lucas-hungryso feel free to cc us if you send it the openstack-dev list (with the right tags)15:42
lucas-hungryHaomeng, great15:42
Haomenglucas-hungry: will read the meeting minutes, tomorrow:)15:43
lucas-hungrycheers, I'm going to grab some food15:43
Haomenglucas-hungry: sure15:43
lucas-hungryHaomeng, g'night :)15:43
Haomenglucas-hungry: let me sent the mail out before sleeping:)15:43
Haomenglucas-hungry: how do you think about this mail title - [openstack-dev][Ironic][Ceilometer]bp:send-data-to-ceilometer ?15:46
Haomenglucas-hungry: and how do you think should we send to mail list and cc to you and deva and eglynn?15:47
*** vkozhukalov has quit IRC15:50
*** ndipanov has quit IRC15:57
NobodyCamgood morning Ironic15:58
*** jistr has quit IRC15:58
*** ndipanov has joined #openstack-ironic16:11
*** nosnos_ has quit IRC16:14
NobodyCambrb16:15
*** matsuhashi has quit IRC16:15
*** aignatov is now known as aignatov_16:16
*** eglynn-afk has joined #openstack-ironic16:20
max_loburmorning Everyone16:21
NobodyCammorning max_lobur :)16:24
*** jistr has joined #openstack-ironic16:34
*** viktors has left #openstack-ironic16:34
*** viktors has joined #openstack-ironic16:34
lucas-hungryHaomeng, sorry was having lunch, yea that would work great :)16:37
lucas-hungrymorning NobodyCam NobodyCam16:37
*** lucas-hungry is now known as lucasagomes16:37
NobodyCammorning / afternoon lucasagomes16:39
*** viktors has left #openstack-ironic16:42
NobodyCamno Ruby taody?16:45
lucasagomesdidn't see her today, maybe she gets online later on16:51
NobodyCam:-p16:52
NobodyCamhey lucasagomes did you the gate failure for https://review.openstack.org/#/c/6882316:52
* lucasagomes looks16:52
*** lsmola_ has quit IRC16:53
NobodyCamlooks like neutron16:54
lucasagomeshmm yea the tempest test16:54
lucasagomeshmm16:55
lucasagomeslooking at the logs but seems there's a couple of errors (idk if related)16:56
lucasagomesBadRequest: Multiple possible networks found, use a Network ID to be more specific.16:56
*** yuriyz has quit IRC16:57
NobodyCamhumm16:57
lucasagomesyea there's one neutron one here16:57
lucasagomesNeutronClientException: 409-{u'NeutronError': {u'message': u'Unable to complete operation on subnet d92bb011-9a46-4972-aca2-ef08955a4fd8. One or more ports have an IP allocation from this subnet.', u'type': u'SubnetInUse', u'detail': u''}}16:57
lucasagomesI will rerun this tests16:57
lucasagomesseems many things failed16:57
NobodyCamya16:57
lucasagomesre-running, removed the +2 as well until I get the results16:58
NobodyCam:)16:58
* NobodyCam makes another cup of joe16:58
*** hemnafk is now known as hemna_16:59
*** mdurnosvistov_ has quit IRC17:05
*** rloo has joined #openstack-ironic17:07
*** rloo has quit IRC17:07
*** rloo has joined #openstack-ironic17:08
*** rloo has quit IRC17:08
*** rloo_ has joined #openstack-ironic17:08
*** rloo_ has quit IRC17:09
*** rloo has joined #openstack-ironic17:09
*** rloo has quit IRC17:10
*** rloo has joined #openstack-ironic17:11
openstackgerritLucas Alvares Gomes proposed a change to openstack/ironic: Sync db.sqlalchemy code from Oslo  https://review.openstack.org/6900417:11
*** rloo_ has joined #openstack-ironic17:12
*** rloo has quit IRC17:14
*** rloo has joined #openstack-ironic17:14
*** rloo has quit IRC17:14
*** rloo has joined #openstack-ironic17:15
*** rloo has quit IRC17:17
*** rloo has joined #openstack-ironic17:17
*** rloo has quit IRC17:18
*** rloo has joined #openstack-ironic17:18
*** rloo has quit IRC17:20
*** rloo has joined #openstack-ironic17:20
*** matty_dubs is now known as matty_dubs|lunch17:23
lucasagomesNobodyCam, in the tripleo approach17:25
lucasagomeswe are using nova bm to deploy the undercloud image?17:25
lucasagomesand then using ironic to deploy the rest of the undercloud?17:25
NobodyCamlucasagomes: at this point YES17:26
lucasagomesI mean, we are using nova bm in the seed vm17:26
lucasagomesright17:26
lucasagomeswouldn't it be easier to test deploying the undercloud image with ironic?17:26
NobodyCamonce we can deploy then we work us in to seed17:26
lucasagomesI see17:26
devanandamorning, all17:28
max_loburmorning devananda17:28
lucasagomesmorning devananda17:28
lucasagomesmorning max_lobur17:29
NobodyCammorning devananda :)17:29
max_loburdevananda , lucasagomes  and others - https://review.openstack.org/#/c/69135/ threading patch, pls review when you have time17:29
max_loburalso futures are already on their way to master17:29
max_loburhttps://review.openstack.org/#/c/66349/3 Doug have approved the patch17:29
lucasagomesnice! will do17:30
devanandaHaomeng, lucasagomes - it's fine if we need to defer ceilometer integration to Juno cycle17:31
devanandait is a bit premature to expect an integrated project to integrate with us before we also become an integrated project anyway17:31
max_lobur:)17:32
rlooNobodyCam: hi. wrt your question https://review.openstack.org/#/c/68665/1/ironic/drivers/modules/ipmitool.py17:32
devanandaHaomeng: ceilometer shouldn't be using IPMI directly -- i pushed back on this a few months ago on the mailing list17:32
lucasagomesdevananda, :) right I was just investigating the progress (or lack of) that bp17:33
devanandaHaomeng: they were proposing to use ipmitool *locally* to monitor hardware on which a ceilometer agent was running. I suggested they use lm_sensors instead17:33
devanandaHaomeng: and interface with ironic for remote monitoring via IPMI17:33
lucasagomesas I got an email asking me that17:33
devanandaHaomeng: also, enjoy your vacation ;)17:33
rlooNobodyCam. Do you mean add a comment there about all the different types of exceptions that could be raised from the try/except?17:33
*** jcooley_ has joined #openstack-ironic17:34
devanandamax_lobur: thanks, i'll take a look in a bit17:36
max_loburthx17:36
NobodyCamrloo: was just a question I had, all the other doc you added had param and raises lines that one didn't17:36
devanandalucasagomes: re: using ironic in the seed vm, yes it's possible, but there is some extra stuff to make seed work by faking cloud-init data17:36
NobodyCamlooks like it would only be os errors there17:37
devanandalucasagomes: so we opted to start with baking it into the undercloud first, then working on the additional glue to make ironic work in seed17:37
rlooNobodyCam. Good point/question. Someone suggested that it made sense to document the exceptions that were raised by internal methods, but not to document exceptions that were raised by methods called by the internal method. this is a reraise. ? I can add it if you think (as a 'user') that it makes sense to see it.17:38
lucasagomesdevananda, gotcha, I will give it a try17:38
*** max_lobur is now known as max_lobur_afk17:39
max_lobur_afkbrb for a meeting17:39
*** vkozhukalov has joined #openstack-ironic17:40
devanandarloo: AIUI, public methods should document the exceptions they raise and any uncaught exceptions that may be raised by private methods they call17:42
rloodevananda: right, but this isn't a public method.17:42
rloodevananda. So should the non-public method have a raises: Exception for reraising?17:43
devanandarloo: cheezy example: http://paste.openstack.org/show/rV4IQYC9lpzZDcHTTCaA/17:44
*** rloo has quit IRC17:44
*** rloo has joined #openstack-ironic17:44
devanandawoops, skipped a line. def pub(): try: should call both _priv and _other_priv17:45
*** Alexei_987 has quit IRC17:46
devanandahttp://paste.openstack.org/show/1MOVM97sLCLgCvLjDTL3/ is better17:46
rloodevananda. Thx. _priv and _other_priv don't have docstrings. If they did, they'd have a :raises:, right?17:48
devanandaexceptions should be caught at higher layers and handled appropriately. so public methods should document the exceptions they expect to be emitting (even if they're originating in a private method). One can reasonably argue that the public method should catch and reraise the private method's exception, but i would say that that's overkill17:48
devanandarloo: correct17:48
rloodevananda. Ok, now, what if _other_priv() called _priv(), caught/reraised the exception from _priv()?17:49
devanandarloo: doc strings in private methods are less important (but not unimportant) because, well, they're private. presumably they're only being called by that class17:49
devanandaone purpose of these doc strings is to faciltate developers using the methods in our classes and APIs17:50
*** rloo has quit IRC17:51
devanandai'm side tracking17:51
*** rloo has joined #openstack-ironic17:51
rloodevananda. So I added docstrings for private methods, and I guess because of that, NobodyCam is asking (reasonable) whether we should add to the docstring for private method A, the exception that is reraised by private method A.17:51
devanandarloo: it wouldn't hurt :)17:52
rloodevananda. Ok then. And to make sure I understand. If private method A did not reraise Exception-from-calling-B, we would not add a "raises" to the docstring for A, cuz it was B that raised it, not A.17:53
NobodyCamdevananda: you ever seen this? [remote rejected] HEAD -> refs/publish/master/bp/deprecate-baremetal-driver (you are not allowed to upload merges)17:55
*** derekh has quit IRC17:55
devanandaNobodyCam: yes. youve merged rather than rebased17:56
NobodyCamack will redoo17:56
devanandaNobodyCam: careful not to lose your work when you rebase now17:56
NobodyCamyes!!!17:56
devanandaNobodyCam: I'd create a new branch locally, make sure it's up to date, then cherrypick your change set(s) onto it17:57
NobodyCamok17:57
*** mdurnosvistov has joined #openstack-ironic18:01
NobodyCambbt... brb18:01
openstackgerritRuby Loo proposed a change to openstack/ironic: Docstrings for ipmitool  https://review.openstack.org/6866518:02
NobodyCamrloo: :) TY18:03
rlooNobodyCam: yw. Thx for noticing. I 'overdocumented' in another case, so wanted to make sure it was ok ;)18:04
*** harlowja_away is now known as harlowja18:05
* NobodyCam notes he has only ever seen one real case of overdoucmenting... 1 /2 paragraphs per 10 lines of code ... that was many years ago, and was in cobalt18:09
*** matty_dubs|lunch is now known as matty_dubs18:14
*** tatyana has quit IRC18:15
*** martyntaylor has quit IRC18:16
NobodyCampost BBT walkies... bbiafm18:19
*** linggao has joined #openstack-ironic18:20
*** agordeev2 has joined #openstack-ironic18:21
*** jcooley_ has quit IRC18:22
*** jcooley_ has joined #openstack-ironic18:23
*** jistr has quit IRC18:32
devanandaNobodyCam: were you working on another rev of 66461 ?18:38
*** ndipanov is now known as ndipanov_gone18:40
NobodyCamdevananda: yes18:43
devanandaNobodyCam: k k18:44
NobodyCamended up stuck on ghe's comment here: https://review.openstack.org/#/c/66461/1/elements/nova/os-apply-config/etc/nova/nova.conf18:44
NobodyCamlol18:44
*** hstimer has joined #openstack-ironic18:52
* NobodyCam preps for meeting... by getting coffe and bagel18:55
*** marty_ has joined #openstack-ironic18:57
*** marty_ is now known as Guest8544718:57
*** hstimer has quit IRC18:58
* NobodyCam heads over to meetnig channel18:59
*** romcheg1 is now known as romcheg19:00
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 3  https://review.openstack.org/6410819:02
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 1  https://review.openstack.org/6002519:02
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 2  https://review.openstack.org/6233119:02
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 6  https://review.openstack.org/6433619:02
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 5  https://review.openstack.org/6427819:02
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Get rid object model `dict` methods part 4  https://review.openstack.org/6425119:02
*** max_lobur_cell has joined #openstack-ironic19:03
*** hemna has quit IRC19:06
*** vkozhukalov has quit IRC19:08
*** rloo has quit IRC19:15
*** rloo has joined #openstack-ironic19:15
*** hemna has joined #openstack-ironic19:19
*** rloo has quit IRC19:22
*** Haomeng|2 has joined #openstack-ironic19:22
*** rloo has joined #openstack-ironic19:22
*** Haomeng has quit IRC19:25
*** rloo has quit IRC19:25
*** rloo has joined #openstack-ironic19:25
*** rloo has quit IRC19:26
*** rloo has joined #openstack-ironic19:26
*** coolsvap has quit IRC19:33
*** rloo has quit IRC19:34
*** martyntaylor has joined #openstack-ironic19:34
*** rloo has joined #openstack-ironic19:34
*** rloo has joined #openstack-ironic19:35
*** rloo has quit IRC19:35
*** rloo has joined #openstack-ironic19:35
*** epim has joined #openstack-ironic19:35
*** rloo has quit IRC19:35
*** rloo has joined #openstack-ironic19:36
*** martyntaylor has quit IRC19:38
*** max_lobur_cell has quit IRC19:51
*** max_lobur has joined #openstack-ironic19:56
NobodyCamgreat meeting all20:00
* NobodyCam gets mo coffee20:00
devanandalucasagomes: so, in the vein of the threading discussion with max_lobur_afk, krow also brought up that it's fairly standard for APIs to respond 202 merely when they GOT the message20:01
devanandalucasagomes: not as a guarantee that they'll DO the requested action20:01
lucasagomeshmm20:01
lucasagomesso we don't need to validate in a sync way?20:01
devanandaso our RPC call to validate the requests before returning 202? Not necessary from an RFC standpoint20:01
devanandaright20:01
*** rloo_ has left #openstack-ironic20:01
lucasagomesI see20:01
lucasagomeswell... the problem is that, if we accept 202 and the validation fails20:02
devanandait's passed auth and been parsed by the API. we can send the RPC cast and return 20220:02
lucasagomeswe won't know unless we look at the los20:02
matty_dubslucasagomes: So I was mostly just curious, but http://stackoverflow.com/questions/626057/is-it-possible-to-cache-post-methods-in-http has a surprisingly-well-researched answer indicating that you can only cache a POST with a Cache-Control/Expires header that permits it.20:02
lucasagomeslogs20:02
devanandathe user is supposed to watch the resource for status changes20:02
lucasagomesor we have to use the last_error for that validation as well20:02
devanandalucasagomes: the larger context to this discussion was coalescing API requests20:03
devanandalucasagomes: eg, if we get 10 requests to power-on the same node20:03
devanandait's perfectly reasonable for Ironic to coalesce those and only issue one driver.power.set_power_state(POWER_ON) command20:04
lucasagomesmatty_dubs, cheers will take a look20:04
lucasagomesdevananda, hmm, right20:04
devanandamatty_dubs: so what if a client includes taht header "by mistake" ?20:04
lucasagomesbut that might be something wrong, why 10 requests is being issued to power on the same node20:04
matty_dubsdevananda: I think only a server could set that.20:04
lucasagomesI understand the RFC pov20:04
devanandamatty_dubs: the proxy may re-issue POST /v1/nodes/ {some: param} a bunch of times, and create a bunch of arbitrary nodes, when they really only meant to make one20:04
lucasagomesjust think it's not very user-friendly20:04
max_loburlucasagomes: ++20:05
devanandamatty_dubs: you mean the POST response body can only be cached if the server set the appropriate headers20:05
devanandamatty_dubs: my bug report is regarding a proxy caching POST requests :)20:05
devanandaor s/proxy/network hiccup/20:05
*** openstackgerrit has quit IRC20:06
matty_dubsAh, okay. I'm not really arguing we should ignore it, but it sounds like something, somewhere is doing something very broken.20:06
*** openstackgerrit has joined #openstack-ironic20:06
matty_dubsLast time I had an HTTP client re-issue POST requests, I emailed 25k customers 6 times in a row.20:06
devanandathe point is, server-side generation of resource UUIDs as a result of a REST API POST is disconnected from the client and could result in dangling resources taht aren't getting returned to the client20:06
max_loburlucasagomes: devananda are there other benefits from sending 202 without check except simplifying the code20:06
lucasagomesmax_lobur, it will be quicker heh20:07
devanandamatty_dubs: faster response tiem for the client. ability for us to coalesce duplicate requests internally. prevents a potential DDOS attack :)20:07
devanandanot taht there couldn't be others, but it solves one obvious one20:07
devanandaclient issues GET /v1/nodes/{uuid}/validate 1000 x/sec20:08
devanandaPUT /v1/nodes/{uuid}/states/power {ON} is a better example20:09
devanandabut either one could crash the BMC in most servers20:09
max_loburlucasagomes: true, a little quicker - no rpc round trip20:09
devanandagranted, there's a trust requirement in that our API is admin-only20:09
lucasagomesdevananda, yea, that was what i was about to mention20:09
lucasagomesthe ironic api has a strict policy20:09
devanandabut if eg. there's a bug in another layer of the stack (nova? heat? tuskar? etc) which is trusted20:10
matty_dubsAh, okay. Coalescing makes a lot of sense there.20:10
devanandathen ironic suddenly becomes a gateway to crash all your servers20:10
devanandawhich, y'know, we might want to prevent20:10
devananda:)20:10
devanandai may be exagerating slightly, but hopefully the point is clear20:11
lucasagomesyea20:11
NobodyCamironic crash-server-pool does not seem like it would get a lot of use20:11
lucasagomesbut that change also sounds like some optimization20:11
devanandalol20:11
lucasagomeswhich I would push back now20:11
devanandalucasagomes: yes. as far as the RPC optimization around 202's, yea, i agree20:12
devanandai'm much keener on max_lobur's work with threading20:12
lucasagomes+120:12
devananda1 RPC round trip to kick off the work ==> user gets a realistic response (error OR 202)20:12
devanandaI think taht's reasonable20:12
lucasagomesalso, I think that validating before is much more user-friendly (thinking if I was the client here)20:12
lucasagomesyea20:12
devanandabacking up to the POST vs PUT -- lucasagomes, any further thoughts?20:13
*** rloo has quit IRC20:13
lucasagomesdevananda, not in the moment, I prefer to read a bit more about it20:14
devanandaNobodyCam: ever seen hardware proxies re-issue POST requests?20:14
lucasagomessee how other people use it20:14
lucasagomesbecause AFAICT that factory-style we use in our api20:14
*** rloo has joined #openstack-ironic20:14
lucasagomesis pretty in many APIs I've seem20:14
NobodyCamnot proxy but load balancers maybe20:14
*** rloo has quit IRC20:15
*** rloo has joined #openstack-ironic20:15
lucasagomesis pretty used*20:15
lucasagomesor very used :P20:16
*** rloo has quit IRC20:16
devanandalucasagomes: hm, side note - we may want to add a no-cache header to our GET responses, too20:16
*** rloo has joined #openstack-ironic20:16
devanandaeg, GET /v1/nodes/UUID/states20:16
*** rloo has quit IRC20:16
devanandaif there is a caching layer above ironic for any reason, and it caches that, the clients will be mightly confused20:17
*** rloo has joined #openstack-ironic20:17
lucasagomesdevananda, +1!!20:17
*** rloo has quit IRC20:17
lucasagomes /drives may change as well20:17
*** rloo has joined #openstack-ironic20:17
devanandalucasagomes: http://www.ebaytechblog.com/2012/08/20/caching-http-post-requests-and-responses/#.Uua-QDfTnQo20:18
devanandalucasagomes: note the POST body digest section20:18
devanandaor tl;dr - proxies may cache POSTs if the MD5 digest of the body is identical20:18
devananda* and the URI is identical20:19
devanandawhich, in our case, would happen today with node generation20:19
devanandafor N nodes that have the same driver, all the POSTs will look identical20:19
lucasagomesyea :/20:20
NobodyCamso add node UUID to the body?20:20
*** rloo has quit IRC20:20
devanandaNobodyCam: to the URI20:20
lucasagomesdevananda, thanks for the article20:20
*** rloo has joined #openstack-ironic20:20
NobodyCamahh yes20:20
devanandait's the driver_info and port's mac_addresses that will be unique today, but we ought to keep port and node creation semantics equivalent20:21
*** rloo has quit IRC20:21
devanandaso i'd suggest something like ...20:21
*** rloo has joined #openstack-ironic20:21
devanandaPUT /v1/nodes/UUID?param=value&param=value20:21
devanandawith no body20:21
*** rloo has quit IRC20:22
devanandaor, if included in the request, put any JSONEncodedDict objects in the body20:22
*** rloo has joined #openstack-ironic20:22
*** jcooley_ has quit IRC20:22
*** jcooley_ has joined #openstack-ironic20:22
*** rloo has quit IRC20:23
devanandaeg: PUT /v1/nodes/UUID?driver=pxe_ipmi  BODY '{"driver_info": {"ipmi_addres": "1.2.3.4", "ipmi_username": "admin", ...} }'20:23
*** rloo has joined #openstack-ironic20:23
lucasagomes"Configure URLs and patterns in the proxy so that it does not cache if there is a match."20:23
devanandayea, but as it stands today, what if there is a load balancer / upstream proxy / network hiccup which causes a clients' POST request to actually get issued twice20:24
devanandawith the current POST generator, we'll create 2 nodes but the client only gets one back20:25
devanandawith PUT as i'm proposing, we'll only create one node, and even if the client doesn't get anything back, they can GET /v1/nodes/UUID and see if it was successful20:25
devanandait's a different approach. has some merits IMO, but we dont need to decide now20:26
lucasagomesdevananda, +120:26
lucasagomesyea I'd like to read a bit more about it20:26
max_loburthe uuid is going to be generated on the client right?20:27
lucasagomesI will read the article you sent as well20:27
lucasagomesmax_lobur, yes20:27
lucasagomesno20:27
lucasagomesserver will generate it20:27
lucasagomesright now is the ironic api that generates it20:27
devanandamax_lobur: yes20:27
devanandaright now ironic-api generates it. i'm proposing that client generates it20:27
max_loburyep20:27
max_loburgot it20:27
max_loburare there some security concerns for such decision?20:28
max_loburI can't imagine any20:28
lucasagomesdevananda, max_lobur if you submit the uuid in the post20:29
lucasagomesour api will recognize/use it20:29
devanandamax_lobur: nothing i can see. the API will is_valid_uuid() the input20:29
max_loburtrue, like a usual input data validation20:30
devanandaand db has a unique constraint on that field20:30
devanandawhich will, in a crude way, coalesce requests too20:30
openstackgerritMikhail Durnosvistov proposed a change to openstack/ironic: Removes use of timeutils.set_time_override  https://review.openstack.org/6743220:31
lucasagomesdevananda, right and if we changed our client to generate the UUID for us?20:31
devanandalucasagomes: wfiw, i dont think we need this stuff now(). but it's interesting discussion around API issues that folks ARE having with openstack20:31
devanandalucasagomes: sure. that's just providing a friendly client lib :)20:31
lucasagomesdevananda, +220:31
lucasagomesdevananda, yea and that already works in our PAI20:31
lucasagomesAPI*20:31
lucasagomesjust created a node with20:31
lucasagomes| uuid                   | 11111111-1111-1111-1111-111111111111                                    |20:31
lucasagomes:D20:31
devanandaahhhh haha!20:31
devanandanice20:31
max_lobur:D20:32
devanandasure, we aren't listing node.uuid in the internal_attrs list20:32
devanandaso client can set it today20:32
devanandadb/sqla/api.py: L28620:34
lucasagomeshttps://github.com/openstack/ironic/blob/master/ironic/db/sqlalchemy/api.py#L286-L28720:34
devanandaheh20:34
devanandayep20:34
lucasagomes:)20:34
max_loburMoreover, it seems to me that I already wrote a tempest test for chassis - specifying uuid on creation :)20:34
lucasagomesturns out it's a good feature heh20:34
lucasagomeshah20:34
max_loburseems a good way to me too20:34
lucasagomesmax_lobur, hah nice!20:34
lucasagomesI can change our client as well20:35
max_loburso20:35
max_loburwe'll need to change the ironic anyway20:35
max_loburto not to allow it generate uuids20:35
max_loburright?20:35
max_loburto make uuid mandatory20:36
lucasagomesmax_lobur, hmm, thinking20:36
max_loburotherwise some other client may think it's ok to not to pass uuids20:37
max_loburand get into troubles20:37
NobodyCamdevananda: sould I bump noda requirments for ironic client to 0.1.1 from 1.0?20:37
NobodyCams/1.0./0.1.020:37
lucasagomesmax_lobur, true...20:38
devanandaNobodyCam: nah20:38
lucasagomesbut I hmm I kinda like our factory-style20:38
lucasagomesI mean to generate the UUID is not inputed by the client20:38
devanandaNobodyCam: that'll require changes to global-requirements. let's not do that until we're close to landing20:38
max_loburit won't change the user experience20:38
lucasagomesand the proxy seems to me like a specific case, and it's an external problem20:38
max_loburjust will be one more requirements to the cliens20:38
lucasagomesyea20:39
lucasagomescan we talk about it tomorrow? (me gotta run soon)20:39
max_loburbut yea, if we talk about someone who uses ironic through curl …:D20:39
lucasagomesmax_lobur, exactly :P20:39
lucasagomesme20:39
lucasagomesheh20:39
lucasagomessometimes20:39
max_loburhe definitely won't like this feature :D20:39
max_loburgotcha! :D20:39
max_loburlucasagomes: sure20:40
max_loburI also wanted to raise background task cancellation question20:40
max_loburbut if lucasagomes needs to run we can talk tomorrow20:40
lucasagomesmax_lobur, if you don't mind, I will have to leave in 19 minutes (but I was thinking about shower before)20:41
lucasagomesit's my last week here in Brazil so I'm going to catch up with some friends20:41
lucasagomesmax_lobur, it's ok if we talk about it tomrorow? devananda ?20:41
max_loburI'm ok with it20:41
max_lobur:)20:42
*** rloo has quit IRC20:42
lucasagomes:) thanks20:42
*** rloo has joined #openstack-ironic20:42
devanandalucasagomes: yep, good night :)20:44
lucasagomesdevananda, max_lobur right, great follow up!20:44
lucasagomesgood night/afternoon devananda max_lobur20:45
NobodyCamnoght lucasagomes20:45
NobodyCamnight even20:45
lucasagomesNobodyCam, night!20:45
max_loburnight Lucas20:45
NobodyCamand with that I'll brb20:46
*** mrda_away is now known as mrda20:46
max_loburbrb too20:46
*** lucasagomes has quit IRC20:46
*** rloo has quit IRC20:48
openstackgerritRoman Prykhodchenko proposed a change to openstack/ironic: Process public API list as regular expressions  https://review.openstack.org/6662120:53
*** rloo has joined #openstack-ironic20:54
*** rloo__ has joined #openstack-ironic20:55
*** rloo__ has quit IRC20:55
*** rloo__ has joined #openstack-ironic20:56
*** rloo__ has quit IRC20:56
*** rloo__ has joined #openstack-ironic20:57
*** rloo has quit IRC20:59
*** rloo__ has quit IRC21:01
*** rloo has joined #openstack-ironic21:01
*** athomas has quit IRC21:01
*** agordeev2 has quit IRC21:02
openstackgerritdekehn proposed a change to openstack/ironic: Adds Neutron support to Ironic  https://review.openstack.org/6607121:06
NobodyCamdevananda: around to click wip21:25
devanandaNobodyCam: yep21:29
NobodyCamwait no requiremnts21:29
NobodyCamfile21:30
NobodyCamok21:32
NobodyCamrebased and requirments re-readded :-p21:32
*** rloo has quit IRC21:34
*** rloo has joined #openstack-ironic21:34
NobodyCamdkehn: you see mr J's -1 for the last patch?21:37
dkehnNobodyCam: yes, interesting21:37
dkehnNobodyCam: rebase right before I git reviewed21:38
dkehnNobodyCam: I'm going through it again21:38
NobodyCamTypeError: _load_one_plugin() takes exactly 6 arguments (5 given)21:38
NobodyCamlots of those21:39
dkehnNobodyCam: saw that in the cosole out, did something make it into the master21:40
NobodyCamnnot that I saw.. havent dug to deep thou21:41
devanandainteresting21:41
dkehnNobodyCam: well I'm assuming its on my end21:41
devanandathat sounds like an upstream change21:41
devanandalemme test it21:41
* NobodyCam eating BLT21:42
NobodyCam- the T21:42
openstackgerritMatt Wagner proposed a change to openstack/ironic: API: Add sample() method on Node  https://review.openstack.org/6553621:44
*** rloo has quit IRC21:50
dkehnNobodyCam: ok, its broke here after a rm .tox and re-run, so rebasing again21:50
*** rloo has joined #openstack-ironic21:50
dkehnNobodyCam: hmm up to date21:51
dkehndevananda: your testing show anything21:52
openstackgerritRuby Loo proposed a change to openstack/ironic: mock's return value for processutils.ssh_execute  https://review.openstack.org/6947921:52
devanandayour patch worked in my existing .tox21:52
*** rloo has quit IRC21:53
*** jcooley_ has quit IRC21:53
devanandaand i'm fighting pip to install --upgrade anything right now21:53
devanandaprobably issues with my cell phone tether...21:53
*** jcooley_ has joined #openstack-ironic21:53
dkehndevananda: it did in mine as well until I rm -rf .tox and rebuilt21:53
*** rloo has joined #openstack-ironic21:53
devanandacan you test master too?21:55
matty_dubsNobodyCam: Oh, I saw those on my patch just now, too21:56
NobodyCammatty_dubs: ??? patches?21:56
matty_dubsNobodyCam: Err, sorry. Just submitted https://review.openstack.org/#/c/65536/ again and Jenkins shot it down with the _load_one_plugin() error you mentioned21:57
NobodyCamahh :)21:58
devanandawer'e currently pinned to stevedore >= 0.12 and latest is 0.1421:58
devanandareleased today21:58
devanandaso i think stevedore 0.14 just broke us -- that's my guess21:58
devanandai'm having trouble testing due to bad connection tho21:58
NobodyCamoh happy happy joy joy21:59
dkehnI am clone a fresh batch and trying22:00
devanandaalso guys, we now have tempest api's in our gate22:01
devanandathat patch finally landed22:01
*** jbjohnso has quit IRC22:01
NobodyCamdevananda: just asking, do you recall why line #158 is in this file? https://review.openstack.org/#/c/66461/1/elements/nova/os-apply-config/etc/nova/nova.conf22:03
NobodyCamnova shouldn't need to know the tftp server , if my thought process is correct22:04
devanandaNobodyCam: you're correct22:05
devanandaNobodyCam: that may have been accidentally left over, or part of the plan to have nova do the initial neutron config. either way it should go22:05
NobodyCam:) will do22:05
NobodyCamjust checking b4 I wack'd it22:06
*** zenfish has joined #openstack-ironic22:20
*** linggao has quit IRC22:20
NobodyCamdevananda: (anyone) off the top of your head dose line #7 loo valid for a conf file?22:21
NobodyCamhttp://paste.openstack.org/show/A5RcoB4qWTRKreLgzx4v/22:21
dkehnNobodyCam: devananda ok llok clone break, is there a fix22:23
dkehnsorry that didn't make much sense, the clone pull  from master broke, is there a resolution yet22:24
NobodyCamlol22:24
NobodyCamnot just yet. devananda was looking at stevedore22:24
dkehnok, thx22:25
*** zenfish has left #openstack-ironic22:26
devanandadkehn: ack, thanks. working on it22:27
*** zenfish has joined #openstack-ironic22:27
*** max_lobur has quit IRC22:40
*** epim has quit IRC22:41
*** jdob has quit IRC22:42
devanandadefinitelyu confirmed - this breaks as soon as i upgrade stevedore to 0.1.422:49
devananda0.1422:49
NobodyCamquick walkies22:53
*** mdurnosvistov has quit IRC23:00
openstackgerritDevananda van der Veen proposed a change to openstack/ironic: Add missing parameter in call to _load_one_plugin  https://review.openstack.org/6949523:06
devanandadkehn, NobodyCam -- in theory, that ^ should fix our gate23:06
devanandawe'll find out in a bit23:06
devanandadkehn: you'll need to rebase on top of that (assuming it passes jenkins)23:07
devanandaoh23:07
devanandaandi need to pin stevedore now23:07
openstackgerritDevananda van der Veen proposed a change to openstack/ironic: Add missing parameter in call to _load_one_plugin  https://review.openstack.org/6949523:08
NobodyCamhumm trying to think of the best way to explose more then one cpu_arch to nova23:10
*** matty_dubs is now known as matty_dubs|gone23:13
devanandahrmm23:17
devanandaNobodyCam: so taht used to be a property of the n-cpu config23:17
devanandaNobodyCam: now it's coming from ironic.node.properties, ya?23:18
NobodyCamya23:18
NobodyCambut I wanted to expose i386 and amd64, but not say arm23:18
devanandawhy?23:19
NobodyCamwhy not expose? or why more then one?23:19
*** Haomeng|2 has quit IRC23:19
devanandaif all the images are the right arch, will the pxe driver run into a problem with arm?23:19
*** Haomeng has joined #openstack-ironic23:20
* NobodyCam notes arm was just a off the cuff example23:20
NobodyCami was wanting the admin to have to set all the supported arches23:21
devanandanode.properties needs to get matched by our nova scheduler plugin to the image and flaor specs23:21
devanandaso23:21
devanandayes, the admin has to set the supported arches by specifying them on teh node's properties AND the images that are loaded into glance23:22
NobodyCamsomething like cpu_arch: ['i386', 'amd64']23:22
devanandawhere would this be set, and why ?23:22
NobodyCamso admins can define that is supported and so the current instance_type extra_specs requirements still work23:24
devanandaan image is exactly one cpu_arch, and a node supports exactly one cpu_arch -- so if the scheduler matches these, and the admin enrolled them properly, it should work23:24
NobodyCamamd64 will not work if all hardware is i38623:25
NobodyCamor worse kinda work23:25
devanandaerm23:25
devanandayou mean "an amd64 image will not work on i386 hardware"23:25
devanandayes?23:25
NobodyCamyes23:25
devanandaright. so i suggested a solution to that above :)23:25
NobodyCamset on the nodes props23:26
devanandayep23:26
* NobodyCam looks into that23:26
devananda)23:26
devanandaty :)23:26
NobodyCamno no TY23:27
lifelessdevananda: to make sure I understand; you're saying we'll support one and only one arch per node ?23:35
devanandalifeless: at a given moment in time23:35
NobodyCamper node yes, (was how I took it)23:35
devanandalifeless: node.properties is mutable23:35
lifelessdevananda: so what about nodes like amd64 that can run either i386 or amd64 images ?23:35
lifelessdevananda: I'm not saying we have to support it, just seeking clarity23:36
devanandalifeless: we could support that -- but it'll be increased complexity in ironic's REST API, in the nova-ironic driver and the nova scheduler, and any UI on top of them23:36
NobodyCamthe logic was just think of would reject i386 images with only amd64 props set23:37
NobodyCams/think/thinking/23:37
devanandainstead of node.properties {'cpu_arch': 'value'}, that value would need to be a list23:37
devanandait'll also complicate things when we try to expose a search method within properties (eg, "find nodes where cpu_arch == amd64")23:39
NobodyCamthat would require changes to compute_capabilities_filter.py}23:39
devanandaNobodyCam: right23:39
devanandanot just the ironic driver23:39
lifelesshow do hypervisors handle this? e.g. amd64 hypervisors can run i386 image, no ?23:40
devanandalifeless: aiui, nova-baremetal doesn't support what you're asking today, either23:40
devanandalifeless: they're not, afaik, matching instance_type_extra_specs:cpu_arch23:40
devanandaat least they weren't -- i'm fairlys ure baremetal was the first to use that23:40
lifelessok, so they basically assume every hypervisor can run every image23:41
lifelessgoing to be a disaster when arm hits :)23:41
devanandawithin a cell / az, i think so23:41
devanandathat's also a potential (hackish) solution for us -- segregate incompatible cpu arch by cell / az23:42
lifelessok, so I think its pragmatically fine to say one and only one arch23:42
mrdalifeless: per node, or per cell?23:43
devanandait's admin-definable, so presumably they could set arch to "x86" and "arm", then tag both 32 and 64-bit miages with "x86", assuming all their hardware supports both23:43
lifelessmrda: per node23:44
NobodyCamdevananda: that would cool but I think it would break several things (just off the top of my head)23:44
devanandaNobodyCam: like?23:44
NobodyCamhummm23:45
mrdaso, by inference, a cell could support multiple archs.  What if a cell can support multiple archs but only has current capacity for one arch?  How would that affect scheduling?23:45
NobodyCamwell looking at compute_capabilities_filter maybe not what I thought23:46
devanandamrda: scheduler would reject requests for images on an arch with no remaining available nodes23:46
devanandamrda: because compute_capabilities_filter would fail all available nodes23:46
devananda*fail for23:47
mrdadevananda: cool, thanks23:48
NobodyCambrb23:54
*** mrda_ has joined #openstack-ironic23:56
*** mrda_ has left #openstack-ironic23:56
devanandagotta run a few errands, bbiahot23:57
NobodyCam:)23:57
« Sunday, 2014-01-26 Index Tuesday, 2014-01-28 »

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!