| rfchapman | NobodyCam: One thing I noticed is, even though I have one box set up for ubuntu and the other box for fedora. no matter which image i deploy, openstack tries it on both boxes. | 00:01 |
|---|---|---|
| openstackgerrit | Yuiko Takada proposed stackforge/ironic-discoverd: Consider dropping check on power state https://review.openstack.org/160223 | 00:02 |
| rfchapman | I would only expect it to deploy on boxes that have the corasponding deploy images. | 00:02 |
| NobodyCam | rfchapman: thats strange. do you have the retry schudeler loded in nova | 00:03 |
| rfchapman | NobodyCam: Not sure, what is it? | 00:03 |
| NobodyCam | http://docs.openstack.org/icehouse/config-reference/content/section_compute-scheduler.html | 00:06 |
| openstackgerrit | Yuiko Takada proposed stackforge/ironic-discoverd: Add scripts to manage translations https://review.openstack.org/158981 | 00:06 |
| *** rloo has quit IRC | 00:12 | |
| *** rloo has joined #openstack-ironic | 00:12 | |
| *** chlong has quit IRC | 00:15 | |
| rfchapman | NobodyCam: it's currently configured as default #scheduler_driver=nova.scheduler.filter_scheduler.FilterScheduler | 00:19 |
| *** hj-hp has quit IRC | 00:22 | |
| rfchapman | OK, I'm starting to see how this functions... I don't have any of the filter/instance types configured. That would explaine why it tries them all. | 00:23 |
| *** mtanino_ has joined #openstack-ironic | 00:28 | |
| *** mtanino has quit IRC | 00:29 | |
| *** BadCub has quit IRC | 00:31 | |
| rfchapman | NobodyCam: When I deploy the fedora image, it seems to give me the ubuntu deploy_kernel I did a md5sum on the image that I created for ubuntu as well as the "deploy_kernel" in the pxe cfg dir and they are the same. | 00:35 |
| NobodyCam | rfchapman: how did you build your deploy ramdisk? | 00:36 |
| *** naohirot has joined #openstack-ironic | 00:46 | |
| rfchapman | NobodyCam: I followed the baremetal instructions. using the diskimager tool | 00:49 |
| rfchapman | I'm sorry the diskimage builder tool | 00:50 |
| NobodyCam | rfchapman: so I expect you only have one type of deployment ramdisk | 00:50 |
| NobodyCam | diskimage-builder/bin/ramdisk-image-create -a amd64 ubuntu deploy-ironic -o <DEPLOY_IMAGE_NAME> <- for ubuntu | 00:50 |
| NobodyCam | diskimage-builder/bin/ramdisk-image-create -a amd64 fedora deploy-ironic -o <DEPLOY_IMAGE_NAME> <- for fedora | 00:51 |
| rfchapman | Yep that is the one. | 00:52 |
| rfchapman | I think I'm going to wipe what I have and start from scratch. I'm thinking I may have made a mistake before. | 00:54 |
| jiangfei|3 | NobodyCam: hi, i have one question. Juno's known issues about ' IPMI passwords are visible to users with cloud admin privileges, via Ironic's API.' This problem has been solved yet? | 00:55 |
| NobodyCam | rfchapman: if you start over may I suggest http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#deploying-ironic-with-devstack | 00:55 |
| rfchapman | Thanks, I'll give it a go. | 00:56 |
| NobodyCam | jiangfei|3: ipmi password are part of driver info | 00:56 |
| NobodyCam | jiangfei|3: I have to check there was some work around that. I'm actually not sure at what state its at | 00:57 |
| jroll | jiangfei|3: they are masked in API responses on the master branch, will be released in kilo | 00:57 |
| jroll | jiangfei|3: per a policy setting IIRC | 00:57 |
| jiangfei|3 | jroll: ok, thank you | 00:58 |
| *** Marga_ has quit IRC | 00:58 | |
| NobodyCam | see :) | 01:01 |
| NobodyCam | thank you jroll | 01:02 |
| * NobodyCam is going to step away. | 01:02 | |
| jiangfei|3 | jroll: does this have a link | 01:02 |
| jroll | jiangfei|3: um, one sec | 01:03 |
| jroll | jiangfei|3: here's the commit, unsure if there's anything else https://github.com/openstack/ironic/commit/efb321c71a709a6f5b33d9de62587117f0c29ba3 | 01:04 |
| jiangfei|3 | jroll: thank you :). | 01:04 |
| jiangfei|3 | jroll: do we need to store ipmitool-password encryption in DB? | 01:11 |
| jroll | jiangfei|3: what do you mean? should ironic encrypt passwords in the database? | 01:11 |
| jiangfei|3 | jroll: yes, currently we store ipmitool-password in DB directlly. we also can get the password by DB operation. so may be we should be encrypted. | 01:14 |
| jroll | jiangfei|3: yes, I think it should, but that's not something we've done yet | 01:16 |
| jroll | jiangfei|3: also, most BMC authentication can be broken if you have network access | 01:16 |
| jiangfei|3 | jroll: yes | 01:17 |
| jroll | we've talked about it, but priorities :( | 01:18 |
| jiangfei|3 | jroll: i got it, i will be merge this commit to my Juno Release, thank you very much. | 01:19 |
| *** krtaylor has quit IRC | 01:20 | |
| jroll | jiangfei|3: you're welcome, don't forget to set policy.json :) | 01:20 |
| jiangfei|3 | jroll: ok | 01:20 |
| mrda | at least a DOS attack on the BMC... | 01:23 |
| openstackgerrit | Merged openstack/ironic: Use mock instead of fixtures when appropriate https://review.openstack.org/159898 | 01:27 |
| *** jmccrory has quit IRC | 01:29 | |
| *** krtaylor has joined #openstack-ironic | 01:33 | |
| *** chenglch|2 has joined #openstack-ironic | 01:34 | |
| *** chenglch|2 has quit IRC | 01:34 | |
| *** chenglch|2 has joined #openstack-ironic | 01:35 | |
| openstackgerrit | Josh Gachnang proposed openstack/ironic: Implement execute clean steps https://review.openstack.org/155561 | 01:37 |
| *** achanda has quit IRC | 01:38 | |
| openstackgerrit | Josh Gachnang proposed openstack/ironic: Fix nits in cleaning https://review.openstack.org/160599 | 01:48 |
| *** rwsu is now known as rwsu-afk | 02:06 | |
| *** dmellado has quit IRC | 02:21 | |
| *** dmellado has joined #openstack-ironic | 02:22 | |
| *** mtanino_ has quit IRC | 02:24 | |
| *** jerryz_ has joined #openstack-ironic | 02:24 | |
| *** lazy_prince has quit IRC | 02:46 | |
| *** yuanying has joined #openstack-ironic | 02:49 | |
| *** yuanying has quit IRC | 02:57 | |
| *** ndipanov_gone has quit IRC | 03:01 | |
| *** yuanying has joined #openstack-ironic | 03:05 | |
| *** ijw has quit IRC | 03:07 | |
| *** yuanying has quit IRC | 03:07 | |
| *** yuanying has joined #openstack-ironic | 03:07 | |
| *** Nisha has joined #openstack-ironic | 03:09 | |
| *** rloo has quit IRC | 03:14 | |
| *** spandhe has quit IRC | 03:16 | |
| *** yuanying has quit IRC | 03:29 | |
| *** yuanying has joined #openstack-ironic | 03:32 | |
| *** yuanying has quit IRC | 03:35 | |
| *** dmellado_ has joined #openstack-ironic | 03:36 | |
| *** Nisha has quit IRC | 03:36 | |
| *** dmellado has quit IRC | 03:37 | |
| *** dmellado_ is now known as dmellado | 03:37 | |
| *** yuanying has joined #openstack-ironic | 03:37 | |
| *** zer0c00l has quit IRC | 03:47 | |
| openstackgerrit | Yuiko Takada proposed stackforge/ironic-discoverd: Add scripts to manage translations https://review.openstack.org/158981 | 03:47 |
| *** zer0c00l has joined #openstack-ironic | 03:48 | |
| *** Marga_ has joined #openstack-ironic | 04:08 | |
| *** harlowja_ is now known as harlowja_away | 04:13 | |
| *** BadCub_Away has joined #openstack-ironic | 04:13 | |
| BadCub_Away | Evening Ironic :-) | 04:16 |
| *** BadCub_Away is now known as BadCub | 04:16 | |
| *** killer_prince has joined #openstack-ironic | 04:18 | |
| *** killer_prince is now known as lazy_prince | 04:18 | |
| *** rameshg87 has joined #openstack-ironic | 04:19 | |
| *** Haomeng has joined #openstack-ironic | 04:21 | |
| *** jmccrory has joined #openstack-ironic | 04:21 | |
| *** jmccrory has joined #openstack-ironic | 04:21 | |
| *** Haomeng|2 has quit IRC | 04:22 | |
| rameshg87 | good morning ironic | 04:29 |
| *** stendulker has joined #openstack-ironic | 04:30 | |
| naohirot | good morning rameshg87 | 04:33 |
| openstackgerrit | Merged stackforge/proliantutils: Support CDROM in get/set persistent boot methods https://review.openstack.org/159799 | 04:33 |
| rameshg87 | naohirot, o/ | 04:34 |
| naohirot | rameshg87: I got different comment about dib vfloppy from dib team. | 04:34 |
| rameshg87 | naohirot, oh let me check | 04:34 |
| rameshg87 | naohirot, they might be making more sense than me :) | 04:35 |
| naohirot | rameshg87: I'd like to know if my patch #2 is different from what you are thinking or not. | 04:35 |
| rameshg87 | naohirot, checking right away .. | 04:35 |
| naohirot | rameshg87: yes, please. | 04:35 |
| *** jmccrory has quit IRC | 04:37 | |
| *** coolsvap_ is now known as coolsvap | 04:37 | |
| rameshg87 | naohirot, yeah this is what i meant | 04:38 |
| rameshg87 | naohirot, i mean have a mechanism for different vendors to plugin their virtual media device filename and their timeouts | 04:38 |
| naohirot | rameshg87: which one? my patch #2 or Clint's comment? | 04:39 |
| rameshg87 | naohirot, patch #2 | 04:39 |
| rameshg87 | naohirot, but i don't quite get Clint's comment | 04:39 |
| rameshg87 | naohirot, we might need to check with him | 04:39 |
| naohirot | rameshg87: Okay, I'll ask Client about his intention. | 04:40 |
| rameshg87 | naohirot, okay | 04:40 |
| rameshg87 | naohirot, so while building the ramdisk, vendors have a mechanism to provide their virtual media device filename, correct ? | 04:41 |
| naohirot | rameshg87: but regarding patch #2, it it the way to passing parameter via kernel parameter? | 04:41 |
| naohirot | rameshg87: Yes, each vendor can pass parameter via env variable. | 04:42 |
| rameshg87 | naohirot, no, in your patch #2 the string will be passed and it will hardcoded in the ramdisk built, right ? | 04:42 |
| rameshg87 | naohirot, if it's urgent for you to get this merged | 04:43 |
| naohirot | rameshg87: Yes, so I meant it is pre-processing when ramdisk is created. | 04:43 |
| rameshg87 | naohirot, i am just wondering if we should go back to your patch set #1 and do this kind of generic mechanism later | 04:43 |
| rameshg87 | naohirot, i am fine doing this later as well .. i understand you need this for kilo | 04:44 |
| naohirot | rameshg87: is patch #2 different from the way you called passing kernel parameter? | 04:44 |
| rameshg87 | naohirot, yeah it | 04:44 |
| rameshg87 | naohirot, yeah it's different, but providing a mechanism for different vendors is all i wanted actually :) | 04:45 |
| rameshg87 | naohirot, right now, only both of us are the consumers | 04:45 |
| *** Nisha has joined #openstack-ironic | 04:46 | |
| naohirot | rameshg87: Is the passing kernel parameter executed in baremetal server? | 04:46 |
| devananda | evening, all | 04:47 |
| rameshg87 | naohirot, actually since the kernel cmdline arguments for the deploy kernel are hardcoded within the iso | 04:47 |
| naohirot | rameshg87: I'm thinking if there is a way to passing parameters not using virtual floppy, | 04:47 |
| rameshg87 | naohirot, we will end up hardcoding the name of virtual media device filename anyway | 04:47 |
| rameshg87 | devananda, o/ | 04:47 |
| naohirot | rameshg87: why do we use virtual flopopy? | 04:47 |
| rameshg87 | naohirot, to pass the arguments for the deploy to the ramdisk | 04:47 |
| *** oomichi_ has joined #openstack-ironic | 04:48 | |
| rameshg87 | naohirot, arguments like iscsi_iqn, ironic_api_url, etc | 04:48 |
| rameshg87 | naohirot, in ilo driver, since we attach the iso file directly from swift, we need to pass the deploy info in some way to the bare metal | 04:48 |
| naohirot | rameshg87: Yeah, I'd like to know how to passing parameter without virtual floppy. | 04:48 |
| rameshg87 | naohirot, i couldn't think of any | 04:49 |
| *** faizan has joined #openstack-ironic | 04:49 | |
| rameshg87 | naohirot, if you can come up with a way - great ;-) | 04:49 |
| naohirot | rameshg87: I'm puzzled :) | 04:49 |
| rameshg87 | :) | 04:50 |
| *** Haomeng has quit IRC | 04:51 | |
| naohirot | rameshg87: So okay, passing parameter via pre-processing the init-func shell script is the way to customize parameter among vendors. Am I right? | 04:51 |
| *** Haomeng has joined #openstack-ironic | 04:51 | |
| rameshg87 | naohirot, yeah | 04:52 |
| naohirot | rameshg87: Yep, thanks. By the way, what 'iirc' stands for? | 04:52 |
| naohirot | rameshg87: I notice other Ironiker used 'iirc' too, today | 04:54 |
| naohirot | rameshg87: s/notice/noticed/ | 04:54 |
| rameshg87 | naohirot, if i remember correctly (iirc) | 04:54 |
| rameshg87 | naohirot, that's way in which you can tell you are not sure about something :) | 04:54 |
| jroll | \o | 04:55 |
| rameshg87 | jroll, o/ | 04:55 |
| naohirot | good evening devananda | 04:55 |
| naohirot | good evening jroll | 04:55 |
| jroll | hi :) | 04:55 |
| jroll | what patch are y'all talking about? | 04:55 |
| rameshg87 | jroll, if the question was to us - it was https://review.openstack.org/#/c/159888/ | 04:56 |
| naohirot | rameshg87: that means "./imagebuild/coreos/full_trusty_build.sh iirc", in this case iirc is not argument? | 04:56 |
| rameshg87 | naohirot, oh | 04:57 |
| rameshg87 | naohirot, i mean if i remember correctly, i ran the script "./imagebuild/coreos/full_trusty_build.sh" | 04:57 |
| naohirot | rameshg87: just IRC abbribiation? | 04:57 |
| rameshg87 | naohirot, yeah | 04:58 |
| naohirot | rameshg87: lol | 04:58 |
| jroll | :P | 04:58 |
| naohirot | rameshg87: Now I got it, If I Remember Correctly = iirc | 04:58 |
| rameshg87 | yeah | 04:59 |
| naohirot | rameshg87: :) | 04:59 |
| naohirot | rameshg87: I thought Ironic something :) | 04:59 |
| jroll | so I think what SpamapS (clint) is saying there is: instead of the __VARIABLE__ placeholders, just use a $VARIABLE, and then set VARIABLE in elements/ramdisk-base/extra-data.d/01-inject-ramdisk-build-files | 04:59 |
| jroll | or rather in that file, write a file with the varaibles, that you can later read the variables from | 05:00 |
| SpamapS | right, basically, rather than edit+parse yourself, just let bash/dash/busybox do your parsing. ;) | 05:00 |
| jroll | :P | 05:01 |
| devananda | meeting time :) | 05:01 |
| jroll | oh snap | 05:01 |
| mrda | \o/ | 05:01 |
| naohirot | jroll: SpamapS: thanks, let me come back after the meeting :) | 05:01 |
| mrda | meeting-3 right? | 05:03 |
| JoshNang | yup | 05:03 |
| mrda | oh good, just others running REALLY late | 05:03 |
| jroll | heh | 05:05 |
| jroll | or not paying attention to the clock >.> | 05:05 |
| *** david-lyle_afk has joined #openstack-ironic | 05:06 | |
| *** pradipta has joined #openstack-ironic | 05:06 | |
| *** stendulker has quit IRC | 05:12 | |
| *** stendulker has joined #openstack-ironic | 05:12 | |
| *** Nisha has quit IRC | 05:16 | |
| *** Nisha has joined #openstack-ironic | 05:16 | |
| *** stendulker has quit IRC | 05:17 | |
| *** stendulker has joined #openstack-ironic | 05:18 | |
| *** stendulker has quit IRC | 05:21 | |
| *** stendulker has joined #openstack-ironic | 05:22 | |
| *** stendulker has quit IRC | 05:22 | |
| *** stendulker has joined #openstack-ironic | 05:22 | |
| *** takadayuiko has joined #openstack-ironic | 05:23 | |
| *** Marga_ has quit IRC | 05:28 | |
| *** lazy_prince has quit IRC | 05:28 | |
| *** killer_prince has joined #openstack-ironic | 05:29 | |
| *** killer_prince is now known as lazy_prince | 05:29 | |
| NobodyCam | can you mount nfs in fuse? | 05:43 |
| *** david-lyle_afk has quit IRC | 05:44 | |
| *** Nisha has quit IRC | 05:45 | |
| *** pradipta has quit IRC | 05:45 | |
| *** Nisha has joined #openstack-ironic | 05:46 | |
| mrda | NobodyCam: I would expect it to be possible | 05:47 |
| mrda | NobodyCam: I just haven't done it | 05:47 |
| rameshg87 | JoshNang, you around ? | 05:51 |
| JoshNang | rameshg87: o/ | 05:51 |
| rameshg87 | JoshNang, before you leave for the day | 05:51 |
| rameshg87 | JoshNang, we will talk on (your tomorrow) | 05:51 |
| *** Nisha_away has joined #openstack-ironic | 05:51 | |
| rameshg87 | JoshNang, regarding the things in zapping :) | 05:51 |
| *** Nisha has quit IRC | 05:51 | |
| JoshNang | sounds good :) i'll pobably be on around 8am pst | 05:52 |
| rameshg87 | JoshNang, sure i will be come around that time too | 05:52 |
| jroll | NobodyCam: I posted a comment on Nisha's thing for you | 05:53 |
| NobodyCam | :) | 05:54 |
| rameshg87 | jroll, NobodyCam, we were talking about that change 2 days back in the channel | 05:55 |
| rameshg87 | jroll, NobodyCam, this was introduced as part of a commit long time back - https://review.openstack.org/#/c/40219/ | 05:55 |
| *** takadayuiko has quit IRC | 05:56 | |
| jroll | rameshg87: indeed | 05:56 |
| rameshg87 | jroll, NobodyCam, wondered why such a check is required in that place | 05:56 |
| rameshg87 | jroll, NobodyCam, i felt it shouldn't be there | 05:56 |
| jroll | rameshg87: I tend to think the same way, but | 05:56 |
| jroll | during inspection the node is locked | 05:57 |
| jroll | but inspection may delete/create ports | 05:57 |
| *** yuanying has quit IRC | 05:58 | |
| rameshg87 | jroll, but the node is locked by the same thread in that place which it has no way to figure out :( | 05:58 |
| jroll | rameshg87: not sure what you mean | 05:59 |
| rameshg87 | jroll, i mean this piece of code: https://github.com/openstack/ironic/blob/master/ironic/db/sqlalchemy/api.py#L136-L144 | 06:00 |
| *** BadCub has quit IRC | 06:00 | |
| jroll | rameshg87: I think it's ok that the node is locked... because that thread has the lock | 06:00 |
| rameshg87 | jroll, currently it checks in the db layer if someone has reserved | 06:00 |
| *** Marga_ has joined #openstack-ironic | 06:01 | |
| jroll | right... | 06:01 |
| rameshg87 | jroll, but if someone locked the node themselves and tried destroy_port | 06:01 |
| rameshg87 | jroll, i assume that would fail | 06:01 |
| rameshg87 | jroll, but have to check how ironic port-delete works :) | 06:02 |
| jroll | rameshg87: right, that's what should happen | 06:02 |
| jroll | OH | 06:02 |
| jroll | rameshg87: I see what you mean, ironic port-delete probably doesn't lock | 06:02 |
| jroll | lol | 06:02 |
| rameshg87 | jroll, yeah | 06:02 |
| rameshg87 | jroll, that's even bigger problem :) | 06:02 |
| jroll | rameshg87: so the real fix is for that RPC call to lock | 06:02 |
| naohirot | jroll: devananda: I read both of your comments. | 06:02 |
| *** stendulker_ has joined #openstack-ironic | 06:02 | |
| jroll | rameshg87: and remove this check | 06:03 |
| rameshg87 | jroll, yeah | 06:03 |
| rameshg87 | jroll, exacly | 06:03 |
| rameshg87 | jroll, we shouldn't be doing any sort of that checks in the db layer | 06:03 |
| devananda | no. we need to | 06:03 |
| devananda | the locking IS done at the db layer | 06:03 |
| jroll | rameshg87: oh, it doesn't get to the API https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/port.py#L355 | 06:03 |
| jroll | devananda: we sholdn't be checking `node.reservation is not None` at the dbapi layer | 06:04 |
| jroll | devananda: we should be acquiring a lock and calling dbapi.destroy_port() | 06:04 |
| rameshg87 | jroll, yeah that's what i meant. calling destroy() without a lock | 06:04 |
| devananda | jroll: right | 06:04 |
| *** stendulker has quit IRC | 06:04 | |
| jroll | rameshg87: indeed, that's a problem | 06:04 |
| devananda | the locking is still done at the db layer | 06:04 |
| jroll | devananda: https://github.com/openstack/ironic/blob/master/ironic/db/sqlalchemy/api.py#L136-L144 | 06:05 |
| devananda | the lock is (and should remain) on the node | 06:05 |
| jroll | is the contention here | 06:05 |
| Nisha_away | devananda, i have proposed the following change for above https://review.openstack.org/#/c/151596/24/ironic/db/sqlalchemy/api.py | 06:05 |
| rameshg87 | devananda, but if it's done at db layer, it has no way to figure out who actually locked ? | 06:05 |
| rameshg87 | devananda, like if we ourselves acquired locked on the node, we should be allowed to destroy port, right ? | 06:05 |
| rameshg87 | devananda, but db layer has no way to identify this | 06:05 |
| jroll | rameshg87: +1 | 06:05 |
| Nisha_away | devananda, but inspection we need to delete the port if MAC is not discovered anymore | 06:05 |
| rameshg87 | Nisha_away, yeah we were talking about that | 06:06 |
| jroll | rameshg87: you should comment about this on gerrit since you're clearly smarter than I am | 06:06 |
| jroll | :) | 06:06 |
| rameshg87 | :) | 06:06 |
| Nisha_away | rameshg87, yes i know | 06:06 |
| naohirot | devananda: jroll: I'll update the spec. This is just my misunderstanding. I didn't take this is a contention, but just suggestion. | 06:06 |
| rameshg87 | jroll, but the problem was we didn't know why that change was made. there was no bug as well :( | 06:07 |
| rameshg87 | jroll, may be code has changed so much after that - it was needed then | 06:07 |
| jroll | rameshg87: I'm finding that with many old changes these days | 06:07 |
| jroll | naohirot: I'm sorry that wasn't clear, however I think it is a real issue | 06:07 |
| naohirot | jroll: I'll follow the Ironic design policy :) | 06:08 |
| Nisha_away | jroll, i saw the comment the db change | 06:08 |
| Nisha_away | why we shouldnt do it based on state? | 06:08 |
| jroll | naohirot: I wouldn't say it's Ironic's policy specifically, but at any rate thank you :) | 06:08 |
| devananda | rameshg87: my suggestion - file a bug describing this, then propose a fix, and base your secure boot patch on that | 06:08 |
| jroll | Nisha_away: I just feel like there's a layer violation there, I think rameshg87's idea is the best option here | 06:09 |
| rameshg87 | devananda, yeah it should be fixed separately | 06:09 |
| rameshg87 | devananda, and it's for ilo inspection (not secure boot) :) | 06:09 |
| devananda | rameshg87: d'oh! yes - i just closed some tabs and looked at the wrong one | 06:10 |
| rameshg87 | :) | 06:10 |
| *** killer_prince has joined #openstack-ironic | 06:10 | |
| jroll | I always do the opposite, open 10 tabs and lose them | 06:10 |
| *** lazy_prince has quit IRC | 06:10 | |
| *** killer_prince is now known as lazy_prince | 06:10 | |
| *** rameshg87 is now known as rameshg87-away | 06:10 | |
| jroll | ok, I think this is a good point for me to go to be | 06:10 |
| jroll | d | 06:10 |
| jroll | especially when I can't type the word bed | 06:11 |
| devananda | ok folks, i'm going to wander off to sleep soon | 06:11 |
| jroll | good night everybody | 06:11 |
| devananda | heh ... yeap, that's a good sign | 06:11 |
| devananda | g'night all | 06:11 |
| Nisha_away | g'night | 06:11 |
| *** achanda has joined #openstack-ironic | 06:14 | |
| naohirot | jroll: no problem, thanks :) | 06:14 |
| *** yuanying has joined #openstack-ironic | 06:16 | |
| naohirot | jroll: anyway I'm still leaning how to develop OSS anyway. I definitly leaned one thing this time :) | 06:16 |
| *** yuanying has quit IRC | 06:17 | |
| naohirot | jroll: s/leaning/learning/ | 06:17 |
| *** yuanying has joined #openstack-ironic | 06:31 | |
| *** russell_h has quit IRC | 06:38 | |
| *** stendulker has joined #openstack-ironic | 06:43 | |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic: follow-up patch for generic node inspection https://review.openstack.org/160665 | 06:44 |
| *** david-lyle_afk has joined #openstack-ironic | 06:44 | |
| *** ukalifon1 has joined #openstack-ironic | 06:45 | |
| *** stendulker_ has quit IRC | 06:46 | |
| *** yuanying has quit IRC | 06:48 | |
| *** yuanying has joined #openstack-ironic | 06:48 | |
| *** Nisha_away has quit IRC | 06:48 | |
| *** jroll has quit IRC | 06:49 | |
| *** yuanying has quit IRC | 06:50 | |
| *** yuanying has joined #openstack-ironic | 06:51 | |
| *** jroll has joined #openstack-ironic | 06:59 | |
| *** yuanying has quit IRC | 07:01 | |
| *** mrda is now known as mrda-away | 07:04 | |
| *** oomichi_ has quit IRC | 07:18 | |
| *** foexle has joined #openstack-ironic | 07:20 | |
| *** openstackgerrit has quit IRC | 07:22 | |
| *** openstackgerrit has joined #openstack-ironic | 07:22 | |
| *** AnxiousGarlic has joined #openstack-ironic | 07:24 | |
| *** AnxiousGarlic has left #openstack-ironic | 07:25 | |
| *** yuanying has joined #openstack-ironic | 07:33 | |
| *** ndipanov_gone has joined #openstack-ironic | 07:35 | |
| *** yuanying has quit IRC | 07:36 | |
| *** Marga_ has quit IRC | 07:48 | |
| *** yuanying has joined #openstack-ironic | 07:54 | |
| *** dlpartain has joined #openstack-ironic | 07:56 | |
| *** krtaylor has quit IRC | 07:58 | |
| *** Marga_ has joined #openstack-ironic | 07:58 | |
| *** ndipanov_gone has quit IRC | 07:59 | |
| *** dlpartain has left #openstack-ironic | 07:59 | |
| *** Haomeng has quit IRC | 08:01 | |
| *** ifarkas has joined #openstack-ironic | 08:01 | |
| *** Haomeng has joined #openstack-ironic | 08:03 | |
| *** Marga_ has quit IRC | 08:06 | |
| *** jcoufal has joined #openstack-ironic | 08:07 | |
| *** krtaylor has joined #openstack-ironic | 08:08 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic: Common changes for secure boot support https://review.openstack.org/153974 | 08:16 |
| *** lsmola has joined #openstack-ironic | 08:16 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic: Secure boot support for pxe_ilo driver https://review.openstack.org/154808 | 08:21 |
| *** mtanino has joined #openstack-ironic | 08:21 | |
| *** yuanying has quit IRC | 08:24 | |
| *** mtanino has quit IRC | 08:26 | |
| *** eglute has quit IRC | 08:26 | |
| *** yuanying has joined #openstack-ironic | 08:26 | |
| *** athomas has joined #openstack-ironic | 08:27 | |
| *** dtantsur|afk is now known as dtantsur | 08:30 | |
| dtantsur | Morning | 08:30 |
| *** yuanying has quit IRC | 08:31 | |
| *** eglute has joined #openstack-ironic | 08:31 | |
| *** chenglch has joined #openstack-ironic | 08:31 | |
| *** chenglch|2 has quit IRC | 08:34 | |
| rameshg87-away | dtantsur, o/ | 08:41 |
| *** rameshg87-away is now known as rameshg87 | 08:41 | |
| dtantsur | o/ | 08:41 |
| openstackgerrit | Merged stackforge/ironic-discoverd: Consider dropping check on power state https://review.openstack.org/160223 | 08:44 |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic: Secure boot support for iscsi_ilo driver https://review.openstack.org/154814 | 08:45 |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic: Secure boot support for agent_ilo driver https://review.openstack.org/154816 | 08:45 |
| *** jistr has joined #openstack-ironic | 08:46 | |
| *** achanda has quit IRC | 08:55 | |
| openstackgerrit | Shivanand Tendulker proposed openstack/ironic: Ilo drivers sets capabilities:boot_mode in node https://review.openstack.org/155731 | 08:57 |
| *** lifeless has quit IRC | 09:05 | |
| *** lifeless has joined #openstack-ironic | 09:06 | |
| *** mgoddard has joined #openstack-ironic | 09:06 | |
| *** andreykurilin_ has joined #openstack-ironic | 09:06 | |
| *** JoshNang has quit IRC | 09:07 | |
| *** MattMan has joined #openstack-ironic | 09:11 | |
| *** andreykurilin_ has quit IRC | 09:12 | |
| *** andreykurilin_ has joined #openstack-ironic | 09:12 | |
| *** vdrok_afk has quit IRC | 09:17 | |
| *** andreykurilin_ has quit IRC | 09:26 | |
| *** rameshg87 is now known as rameshg87-brb | 09:27 | |
| *** andreykurilin_ has joined #openstack-ironic | 09:27 | |
| *** Haomeng|2 has joined #openstack-ironic | 09:29 | |
| *** Haomeng has quit IRC | 09:30 | |
| *** romcheg has joined #openstack-ironic | 09:30 | |
| *** vdrok has joined #openstack-ironic | 09:32 | |
| vdrok | morning ironic | 09:32 |
| vdrok | dtantsur, morning, around? | 09:33 |
| dtantsur | vdrok, morning yeah | 09:34 |
| vdrok | dtantsur, have a question about your comment here - https://review.openstack.org/#/c/136741/31/doc/source/deploy/install-guide.rst | 09:34 |
| *** romcheg has quit IRC | 09:35 | |
| vdrok | dtantsur, when I tested it without dhcp provider everything worked fine without additional config editing | 09:35 |
| openstackgerrit | Yuriy Zveryanskyy proposed openstack/ironic: Do not save auth token on TFTP server in PXE driver https://review.openstack.org/159819 | 09:35 |
| vdrok | dtantsur, switch_pxe_config seems to work fine | 09:35 |
| dtantsur | well, maybe, I didn't test it personally | 09:35 |
| vdrok | dtantsur, ok, then I'll just mention local boot :) | 09:36 |
| dtantsur | vdrok, it's worth mentioning :) if you can confirm things work with PXE boot too, that's good :) | 09:36 |
| vdrok | dtantsur, will do | 09:36 |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic: Follow-up patch for generic node inspection https://review.openstack.org/160665 | 09:39 |
| *** Nisha has joined #openstack-ironic | 09:41 | |
| Nisha | dtantsur, hi | 09:43 |
| dtantsur | o/ | 09:44 |
| Nisha | just responded to one of the comment in follow up patch | 09:44 |
| Nisha | did u happen to see it? | 09:44 |
| dtantsur | looking now | 09:45 |
| *** romcheg has joined #openstack-ironic | 09:46 | |
| dtantsur | Nisha, my only point is that fallback last_error makes no sense whatsoever and should be deleted :) either assume last_error is always provided or use some sane default | 09:46 |
| Nisha | i just raised the patch without addressing it | 09:46 |
| Nisha | but if we use timeout word ruby has objection | 09:46 |
| Nisha | she is like this method can be used for any filter criteria | 09:47 |
| Nisha | it may not be timeout always | 09:47 |
| *** naohirot has quit IRC | 09:49 | |
| Nisha | dtantsur, ^^^^ | 09:50 |
| dtantsur | Nisha, ok, but this wording is even worse, because it lies to a user... please do something about it (e.g. delete completely) | 09:51 |
| *** foexle has quit IRC | 09:51 | |
| *** coolsvap is now known as coolsvap_ | 09:52 | |
| Nisha | is it ok to leave last_erro empty if its not provided? | 09:52 |
| Nisha | if thats fine i will delete it | 09:52 |
| dtantsur | Nisha, I think so | 09:53 |
| dtantsur | Nisha, if folks invent some better wording - good | 09:54 |
| Nisha | dtantsur, so should i wait for comments? | 09:54 |
| Nisha | or should i remove it? | 09:54 |
| dtantsur | Nisha, go ahead :) | 09:54 |
| Nisha | i will reove then | 09:55 |
| Nisha | remove* | 09:55 |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic: Follow-up patch for generic node inspection https://review.openstack.org/160665 | 09:59 |
| Nisha | dtantsur, ^^^ removed the default error message | 09:59 |
| *** foexle has joined #openstack-ironic | 10:08 | |
| openstackgerrit | Ramakrishnan G proposed stackforge/proliantutils: HPSSA: Handle all storage units https://review.openstack.org/159001 | 10:26 |
| *** PaulCzar has quit IRC | 10:27 | |
| *** dtantsur is now known as dtantsur|bbl | 10:28 | |
| *** andreykurilin_ has quit IRC | 10:46 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Check UUID correctness for Glance images https://review.openstack.org/151951 | 10:49 |
| *** chenglch has quit IRC | 10:50 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Address comments on Iec8e1f862d29639363b71c67d60e711d35d2ed94 https://review.openstack.org/160732 | 10:51 |
| *** pelix has joined #openstack-ironic | 11:03 | |
| *** Haomeng has joined #openstack-ironic | 11:05 | |
| *** rameshg87-brb has quit IRC | 11:06 | |
| *** Haomeng|2 has quit IRC | 11:06 | |
| *** stendulker has quit IRC | 11:13 | |
| *** coolsvap_ is now known as coolsvap | 11:17 | |
| *** Nisha has quit IRC | 11:35 | |
| *** Nisha_away has joined #openstack-ironic | 11:35 | |
| *** romcheg has quit IRC | 11:49 | |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic: Follow-up patch for generic node inspection https://review.openstack.org/160665 | 11:58 |
| openstackgerrit | Tan Lin proposed openstack/ironic: Add server's supported API versions to exception HTTPNotAcceptable(406) https://review.openstack.org/160758 | 12:02 |
| *** EmilienM|afk is now known as EmilienM | 12:05 | |
| *** killer_prince has joined #openstack-ironic | 12:06 | |
| *** lazy_prince has quit IRC | 12:07 | |
| *** killer_prince is now known as lazy_prince | 12:07 | |
| openstackgerrit | Naohiro Tamura proposed openstack/ironic-specs: iRMC Virtual Media Deploy Driver for Ironic https://review.openstack.org/134865 | 12:20 |
| *** dtantsur|bbl is now known as dtantsur | 12:22 | |
| *** gridinv has joined #openstack-ironic | 12:33 | |
| *** faizan has quit IRC | 12:37 | |
| openstackgerrit | Imre Farkas proposed openstack/ironic: POC: DRAC RAID configuration https://review.openstack.org/145464 | 13:02 |
| *** jroll has quit IRC | 13:07 | |
| *** jroll has joined #openstack-ironic | 13:07 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: Add iter_nodes() helper to the conductor manager https://review.openstack.org/159100 | 13:08 |
| *** ifarkas has quit IRC | 13:09 | |
| *** ifarkas has joined #openstack-ironic | 13:12 | |
| *** Nisha_away has quit IRC | 13:18 | |
| *** dprince has joined #openstack-ironic | 13:26 | |
| *** ukalifon1 has quit IRC | 13:37 | |
| *** romcheg has joined #openstack-ironic | 13:42 | |
| *** jjohnson2 has joined #openstack-ironic | 13:45 | |
| openstackgerrit | Dmitry Tantsur proposed openstack/ironic: Add module for in-band inspection using ironic-discoverd https://review.openstack.org/156562 | 13:47 |
| *** rameshg87 has joined #openstack-ironic | 13:51 | |
| rameshg87 | good evening folks :) | 13:52 |
| *** rameshg87 is now known as rameshg87-brb | 13:52 | |
| *** trown|outttypeww is now known as trown | 13:54 | |
| *** absubram has joined #openstack-ironic | 13:55 | |
| *** rloo has joined #openstack-ironic | 14:00 | |
| openstackgerrit | Tan Lin proposed openstack/python-ironicclient: Enable ironicclient with --ironic-api-version 1.x https://review.openstack.org/155624 | 14:03 |
| *** kkoski has joined #openstack-ironic | 14:15 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Fix nits for supporting non-glance images https://review.openstack.org/160732 | 14:16 |
| openstackgerrit | Nisha Agarwal proposed openstack/ironic: Follow-up patch for generic node inspection https://review.openstack.org/160665 | 14:18 |
| *** lazy_prince has quit IRC | 14:19 | |
| jroll | morning all :) | 14:19 |
| *** kkoski has quit IRC | 14:20 | |
| dtantsur | jroll, morning! | 14:21 |
| *** killer_prince has joined #openstack-ironic | 14:24 | |
| *** killer_prince is now known as lazy_prince | 14:24 | |
| *** romcheg has quit IRC | 14:24 | |
| *** kkoski has joined #openstack-ironic | 14:24 | |
| openstackgerrit | Naohiro Tamura proposed openstack/ironic-python-agent: Add iRMC virtual floppy media support https://review.openstack.org/160813 | 14:25 |
| jroll | dtantsur: thoughts on getting a spec for this? https://review.openstack.org/#/c/159819/5 | 14:26 |
| jroll | there's a lot to bikeshed about there | 14:26 |
| dtantsur | well... | 14:26 |
| *** kkoski has quit IRC | 14:26 | |
| *** erwan_taf has quit IRC | 14:26 | |
| dtantsur | jroll, not sure, what are your bikeshed points? my only concern is doing it that late in the cycle | 14:27 |
| *** chlong has joined #openstack-ironic | 14:27 | |
| jroll | well, that's one thing | 14:27 |
| jroll | but this doesn't actually improve security of the deploy | 14:27 |
| jroll | just security of the admin creds | 14:28 |
| *** absubram has quit IRC | 14:28 | |
| jroll | that is to say, I'm +1 on that patch but I think there's much more to do | 14:28 |
| dtantsur | I personally treat it like an important improvements | 14:28 |
| *** kkoski has joined #openstack-ironic | 14:28 | |
| dtantsur | (and matches what you do with the agent and I do with discoverd) | 14:28 |
| *** kkoski has quit IRC | 14:29 | |
| *** kkoski has joined #openstack-ironic | 14:29 | |
| jroll | yeah | 14:29 |
| jroll | ok, sounds good :) thanks | 14:30 |
| * jroll also noticed when looking at test failures that we should set trace mode in our bash ramdisk | 14:30 | |
| *** romcheg has joined #openstack-ironic | 14:30 | |
| *** Marga_ has joined #openstack-ironic | 14:35 | |
| *** BadCub has joined #openstack-ironic | 14:36 | |
| BadCub | Morning Ironic | 14:36 |
| *** rameshg87-brb is now known as rameshg87 | 14:37 | |
| *** Marga_ has quit IRC | 14:37 | |
| jroll | morning BadCub :) | 14:38 |
| *** Marga_ has joined #openstack-ironic | 14:38 | |
| *** ndipanov has joined #openstack-ironic | 14:38 | |
| rameshg87 | question to everyone: why do we have this kind of checking in our api: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L66-L90 | 14:38 |
| rameshg87 | do we intend to add such kind of checks (in the future) for disallowing features based on minor version of api ? | 14:39 |
| jroll | rameshg87: exactly that | 14:39 |
| rameshg87 | jroll, but it will be headache, right ? | 14:40 |
| jroll | we need to make it better so we don't clutter the code | 14:40 |
| dtantsur | it already is :D | 14:40 |
| rameshg87 | jroll, to add a method of disallowing everytime | 14:40 |
| jroll | for us, yes, for deployers it will help them (I think) | 14:40 |
| jroll | right... | 14:40 |
| rameshg87 | jroll, how ? | 14:40 |
| jroll | so nova does some decorator madness | 14:40 |
| jroll | rameshg87: so tooling doesn't break because we change the API | 14:40 |
| rameshg87 | jroll, sorry, what is tooling here ? | 14:40 |
| jroll | rameshg87: tools that operators use to interact with the API | 14:41 |
| rameshg87 | jroll, to bundle appropriate version of client or something ? | 14:41 |
| jroll | well, the client should have an argument for the version | 14:41 |
| jroll | so a good example is when we changed NOSTATE to AVAILABLE | 14:41 |
| jroll | without this check, my tools would break: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L59-63 | 14:42 |
| jroll | (also nova and tempest would break) | 14:42 |
| rameshg87 | jroll, i agree for that | 14:42 |
| rameshg87 | jroll, but why for logical names ? | 14:42 |
| rameshg87 | jroll, logical names was a new feature right ? | 14:42 |
| * dtantsur is not sure too... | 14:42 | |
| rameshg87 | jroll, no client before that would know that we can pass logical names in /v1/nodes/<>/ | 14:42 |
| rameshg87 | jroll, previous clients would keep on sending uuid and we are still happy with it | 14:43 |
| jroll | rameshg87: the client is released separately fromt the server... | 14:43 |
| *** zz_jgrimm is now known as jgrimm | 14:43 | |
| jroll | my client may be newer than my server | 14:43 |
| *** coolsvap is now known as coolsvap_ | 14:43 | |
| rameshg87 | jroll, oh so we provide support for older client with newer server ? | 14:44 |
| rameshg87 | jroll, i don't think that's possible :) | 14:44 |
| jroll | I don't see why not | 14:44 |
| rameshg87 | jroll, i add 3 new api endpoints for raid | 14:44 |
| rameshg87 | jroll, newer clients know about it | 14:44 |
| rameshg87 | jroll, but older server doesn't know about it | 14:44 |
| rameshg87 | jroll, so newer client + older server is broken, right ? | 14:44 |
| jroll | rameshg87: it's not broken, it just doesn't support raid | 14:45 |
| rameshg87 | jroll, hmm .. | 14:45 |
| jroll | rameshg87: also, keep in mind some tols may not use the client | 14:46 |
| jroll | rameshg87: bumping the version for a new feature makes that feature discoverable | 14:46 |
| rameshg87 | jroll, do you direct curl requests ? | 14:46 |
| jroll | the user can know if logical names is supported just by checking if v1.5 works | 14:46 |
| jroll | rameshg87: sometimes | 14:47 |
| jroll | rameshg87: and sometimes I use python-request | 14:47 |
| jroll | rameshg87: and sometimes I use node.js | 14:47 |
| rameshg87 | jroll, bumping the version i agree | 14:47 |
| rameshg87 | jroll, but separate disallowing checks based on minor version, why would we do it for features ? | 14:48 |
| jroll | rameshg87: it makes the feature discoverable without trying the actual endpoint | 14:48 |
| jroll | e.g. I can do /v1/nodes/uuid on v1.5 to see if logical names are available | 14:48 |
| jroll | rather than doing /v1/nodes/name, getting a 404, and wondering if logical names isn't available or if I just got the name wrong | 14:49 |
| rameshg87 | jroll, but isn't there a generic mechanism to query the current max version of api | 14:49 |
| rameshg87 | jroll, if someone knows from 1.5 logical names is supported | 14:50 |
| rameshg87 | jroll, they can just query it and figure out also, right ? | 14:50 |
| * rameshg87 wonders if i am missing something | 14:50 | |
| jroll | hrm | 14:50 |
| rloo | rameshg87: have you seen the nova spec on microversions, that might help | 14:50 |
| jroll | rameshg87: I see your point... however this check is also used for setting a name etc | 14:51 |
| rloo | rameshg87: i can't seem to copy/paste. maybe i should go back to sleep | 14:51 |
| * jroll hands rloo his caffeine IV | 14:51 | |
| rameshg87 | rloo, i haven't. should read it first ? | 14:51 |
| rameshg87 | :) | 14:51 |
| * rameshg87 searches for devanandas mail | 14:52 | |
| rloo | rameshg87: that's what ironic's microversioning is based off of | 14:52 |
| jroll | rameshg87: http://specs.openstack.org/openstack/nova-specs/specs/kilo/approved/api-microversions.html | 14:52 |
| rameshg87 | rloo, jroll, but i still it's odd to add checks for each feature :( | 14:52 |
| * rameshg87 wonders if nova does the same | 14:53 | |
| rloo | rameshg87: to be honest, I haven't totally understood the microversioning too, and when to bump up the microversion etc. | 14:53 |
| jroll | rameshg87: nova is trying to but they are slow, see that spec | 14:53 |
| devananda | so one thing to keep in mind as we continue to evolve the API, even with microversions, we must not break compatibility with older clients within the /v1/ URI | 14:53 |
| rameshg87 | devananda, but for features (like logical names) backward compatibility is not broken | 14:54 |
| devananda | correct | 14:54 |
| jroll | rameshg87: another point -- providing well-defined, versioned, APIs helps us do frequent releases, helps deployers do continuous deploymeny, helps improve our testing story (if we can trust the API to not break, we can do less integration tests) | 14:54 |
| * jroll would like to convince other projects to do this (well) | 14:55 | |
| rameshg87 | devananda, my question was do we still add a check to disallow a particular feature (for minor api versions before it was introduced) ? | 14:55 |
| rameshg87 | devananda, even if it's truly a new feature (i mean an extension which doesn't break backward compatibility) | 14:56 |
| devananda | rameshg87: ah, good point. | 14:56 |
| devananda | in principle, no | 14:56 |
| devananda | however, since we have the facility to do so easily, we will be tempted to do so | 14:56 |
| rameshg87 | oh | 14:56 |
| devananda | I'm not sure that we *should* | 14:57 |
| rameshg87 | devananda, so i was checked for code: https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/node.py#L66-L90 | 14:57 |
| rameshg87 | devananda, do you feel these are not actually required ? | 14:57 |
| rameshg87 | devananda, logical names, driver internal info, manage+inspect | 14:57 |
| rameshg87 | devananda, all these are new things as i understand | 14:57 |
| rloo | rameshg87: the logical names does change the response of eg a node show, because there is now a 'name' field. so even if you use uuids, if we return that new field, it could break someone's whatever if they weren't expecting that new field. | 14:58 |
| openstackgerrit | Imre Farkas proposed openstack/ironic: DRAC RAID testing: provide a way to manipulate target_raid_config from API -- DO NOT MERGE https://review.openstack.org/160825 | 14:58 |
| devananda | rameshg87: hide_driver_internal_info is definitely not required -- an older client should just ignore that part of the response | 14:58 |
| rameshg87 | okay | 14:59 |
| devananda | but rloo is correct re: logical names. that one could easily break an older client in some circumstances (eg, if the client didn't properly type check its input and someone passed it a name) | 14:59 |
| rameshg87 | devananda, but we have a official python-ironicclient for everyone to use | 15:00 |
| jroll | we can't require people to use the official requirement | 15:00 |
| jroll | and it would be silly to do so | 15:00 |
| devananda | rameshg87: sure. but there's also openstackclient (common client library) and other language clients | 15:00 |
| rameshg87 | jroll, i didn't mean that | 15:00 |
| jroll | s/requirement/client/ | 15:01 |
| rameshg87 | devananda, if i understood correctly the concern was | 15:01 |
| devananda | python-ironicclient, while produced by us, is definitely not the only client | 15:01 |
| jroll | rameshg87: I should say, we can't break other clients just because we have the official client | 15:01 |
| devananda | jroll: well said | 15:01 |
| rameshg87 | devananda, someone calling /v1/nodes/my-node would have expected an error if they didn't know about logical names | 15:01 |
| jroll | for example, I have a node.js application that talks to ironic | 15:01 |
| devananda | rameshg87: correct | 15:01 |
| jroll | if you break that I will throw things at you | 15:01 |
| rameshg87 | devananda, but if newer server had such a node, it might return something | 15:02 |
| *** gridinv has quit IRC | 15:02 | |
| devananda | rameshg87: similarly, calling PUT {'target': 'inspect'} /v1/nodes/NNNN/state/provision | 15:02 |
| devananda | rameshg87: or s/inspect/any other random word/ -- someone would expect an error | 15:03 |
| *** gridinv has joined #openstack-ironic | 15:03 | |
| devananda | but we changed what verbs are allowed there, thus we changed what the client should expect | 15:03 |
| rameshg87 | devananda, how about new endpoints ? | 15:03 |
| rameshg87 | devananda, i have in my generic raid code 3 new endpoints for raid: https://review.openstack.org/#/c/155230/ | 15:03 |
| rameshg87 | devananda, should i disallow based on minor version ? | 15:03 |
| rameshg87 | devananda, i would think there is no need | 15:04 |
| NobodyCam | Good Morning Ironic | 15:04 |
| rameshg87 | devananda, someone using older client might never know of such a feature | 15:04 |
| rameshg87 | NobodyCam, o/ | 15:04 |
| NobodyCam | morining devananda rameshg87 | 15:04 |
| NobodyCam | :) | 15:04 |
| devananda | rameshg87: fair point. I'm inclined to agree -- but let me think on it a bit | 15:05 |
| jroll | I'm curious if disallowing a new endpoint/feature/whatever based on the version hurts anything... need to ponder | 15:05 |
| jroll | morning NobodyCam :) | 15:05 |
| rameshg87 | devananda, okay .. | 15:05 |
| NobodyCam | morning jroll :) | 15:05 |
| rameshg87 | jroll, basically for every new feature i feel it just bloats the code | 15:05 |
| devananda | jroll: I can't see it hurting anything to hide it behind a version. but I'm also not sure it /helps/ anything | 15:05 |
| * devananda gets coffee | 15:05 | |
| * NobodyCam seconds the coffee statement | 15:06 | |
| jroll | devananda: yeah | 15:06 |
| jroll | rameshg87: yeah, I hear you | 15:06 |
| rameshg87 | jroll, :) | 15:06 |
| rameshg87 | jroll, thanks :) | 15:06 |
| jroll | rameshg87: so I think we should make the helper code better so that we don't bloat the code :) | 15:07 |
| rameshg87 | jroll, exactly | 15:07 |
| jroll | rameshg87: see how nova is using decorators, that might be a good option | 15:07 |
| rameshg87 | jroll, @disallow_below_api_version(min_version=1.5) | 15:08 |
| rameshg87 | jroll, something like that ? | 15:08 |
| rameshg87 | jroll, if that feature provides a new function in itself | 15:08 |
| jroll | rameshg87: indeed, something like that... what nova is doing is different | 15:08 |
| rloo | jroll, rameshg87: now that my copy/paste is working again. this might be useful too: docs.openstack.org/developer/nova/devref/api_microversions.html | 15:08 |
| jroll | http://specs.openstack.org/openstack/nova-specs/specs/kilo/approved/api-microversions.html#implementation-design-details | 15:08 |
| rameshg87 | jroll, yeah similar thing | 15:09 |
| rameshg87 | jroll, for things like verbs | 15:09 |
| rameshg87 | jroll, rloo, we could just do with method calls | 15:09 |
| rameshg87 | jroll, rloo, i guess anything that would simplify might help | 15:09 |
| jroll | rameshg87: yeah, need to thing/talk about it | 15:10 |
| rameshg87 | jroll, yeah may be live with it for kilo and refactor in early L release ? | 15:10 |
| jroll | rameshg87: yeah, I think so, too much code in flight to be refactoring that stuff around IMO | 15:11 |
| rloo | i was thinking that if we could punt this til L*, we could talk to the nova folks and get their experience | 15:11 |
| rameshg87 | okay | 15:11 |
| *** absubram has joined #openstack-ironic | 15:11 | |
| NobodyCam | rameshg87: jlvillal: thank you for the reviews will address shortly | 15:12 |
| *** Marga_ has quit IRC | 15:13 | |
| *** jerryz_ has quit IRC | 15:15 | |
| rameshg87 | rloo, https://review.openstack.org/#/c/150142/22/ironic/conductor/manager.py | 15:18 |
| rloo | rameshg87: ? | 15:18 |
| rameshg87 | rloo, can't we just a non-shared lock for infer_image_type ? | 15:18 |
| rloo | rameshg87: no | 15:18 |
| rloo | rameshg87: i mean, we could, almost anything is possible. but we shouldn't. | 15:18 |
| rameshg87 | rloo, why ? | 15:18 |
| NobodyCam | morning rloo :) | 15:19 |
| rloo | rameshg87: cuz up to now, the validate calls are meant to validate, not change the node | 15:19 |
| rloo | morning NobodyCam | 15:19 |
| rameshg87 | rloo, but i assume we are not supposed to write to database without having a non-shared lock, right ? | 15:19 |
| rloo | rameshg87: I don't see any reason for that to change for this. i'd rather the code be more inefficient for now, than to add yet another lock. we already have a problem with too many exclusive locks. | 15:19 |
| rameshg87 | rloo, i mean we could degrade the lock after we are done calling infer_image_type | 15:19 |
| rloo | rameshg87: yeah, w/o an exclusive lock, one shouldn't modify a node, but ... | 15:20 |
| rameshg87 | rloo, degrade into a shared lock | 15:20 |
| rloo | rameshg87: what if you can't get an exclusive lock. then what? err out on the validate()? | 15:20 |
| gridinv | jroll: hi, i am looking at network provider bp, trying to implement it for nuage neutron plugin/tor switch. | 15:21 |
| rameshg87 | rloo, but with shared=True would we always succeed ? | 15:21 |
| rameshg87 | rloo, i mean i have to check the code | 15:21 |
| rloo | rameshg87: no, even with shared=true, if a node has been exclusively locked, it would fail | 15:21 |
| jroll | gridinv: hi! how can I help? :) | 15:21 |
| gridinv | jroll: basically got it working, have a issue though | 15:22 |
| dtantsur | rloo, will it? | 15:22 |
| dtantsur | shared lock is not a lock, it's just get | 15:22 |
| rloo | rameshg87: it is a problem today. if you do a validate and say a periodic task is doing a power sync, you could get a fail. | 15:22 |
| rameshg87 | rloo, i would assume 2 people could have locks with shared=True and shared=True | 15:22 |
| NobodyCam | mornign dtantsur :) | 15:22 |
| rameshg87 | rloo, but not with shared=True and shared=False | 15:22 |
| gridinv | jroll: plugin has same code base for ironic/libvirt | 15:22 |
| dtantsur | oh morning NobodyCam, devananda, rloo and other folks :) | 15:22 |
| rloo | dtantsur: can you get a shared lock for the same node that someone has taken an exclusive lock on? | 15:22 |
| NobodyCam | :) | 15:22 |
| rloo | hi dtantsur ;) | 15:22 |
| gridinv | jroll: When nova creates port for instance there is no way to distinguish between those cases | 15:22 |
| rameshg87 | rloo, yeah | 15:22 |
| gridinv | jroll: But in case of ironic, plugin should avoid to do some call to sdn controller. | 15:23 |
| gridinv | jroll: did you ever felt the need to be able to ditinguish between vm/bm cases? | 15:23 |
| dtantsur | rloo, shared lock is not a lock, it's just a get request to database | 15:23 |
| rameshg87 | dtantsur, but nobody is supposed to modify anything while i have shared lock | 15:23 |
| dtantsur | rameshg87, in a non-atomic way - yes | 15:24 |
| jroll | gridinv: so, we solve this by not actually writing the configs until ironic says to write them in https://github.com/rackerlabs/ironic-neutron-plugin | 15:24 |
| dtantsur | rloo, rameshg87, https://github.com/openstack/ironic/blob/master/ironic/conductor/task_manager.py#L197-L200 | 15:24 |
| jroll | gridinv: we have a 'commit' argument that defaults to False, ironic sends commit=True and then the config gets committed to the switch | 15:24 |
| *** anderbubble has joined #openstack-ironic | 15:24 | |
| jroll | gridinv: we also don't run ironic-neutron-plugin for virt things, only baremetal | 15:25 |
| rameshg87 | dtantsur, oh | 15:25 |
| *** alexpilotti has joined #openstack-ironic | 15:25 | |
| jroll | gridinv: to be clear, when I say "we" here I mean rackspace, not ironic | 15:26 |
| dtantsur | rameshg87, rloo, actually, as shared lock is not a lock at all, we should not modify database even atomically while holding only it IMO | 15:26 |
| rloo | dtantsur: thx. wondering now what i was thinking of... | 15:26 |
| rameshg87 | rloo, then i think it nova-validate cannot fail during a power-state sync | 15:26 |
| rameshg87 | s/nova-validate/node-validate | 15:26 |
| rameshg87 | dtantsur, thanks | 15:26 |
| rloo | dtantsur: https://github.com/openstack/ironic/blob/master/ironic/conductor/task_manager.py#L209 | 15:26 |
| rloo | dtantsur: that can happen if the the node wasn't locked | 15:27 |
| dtantsur | well yeah... | 15:27 |
| gridinv | jroll: yes, i had a look at it. wanted to avoid changinng existing plugin data model. | 15:27 |
| rameshg87 | rloo, so rloo it's a change of behaviour, right ? | 15:27 |
| rameshg87 | rloo, that node-validate can fail if we do change to shared=False there | 15:28 |
| rameshg87 | rloo, i mean today it doesn't fail (assuming) | 15:28 |
| *** wuhg has quit IRC | 15:29 | |
| jroll | I believe node-validate does fail if something else has a lock | 15:30 |
| * jroll will check | 15:30 | |
| *** kkoski1 has joined #openstack-ironic | 15:31 | |
| *** kkoski has quit IRC | 15:31 | |
| * rloo wonders how/whether this discussion becomes moot when we implement VALIDAT* states | 15:31 | |
| jroll | oops, rameshg87 is right, it doesn't fail today | 15:32 |
| jroll | VALIDAT* states?? | 15:32 |
| rameshg87 | brb | 15:32 |
| rloo | jroll: http://specs.openstack.org/openstack/ironic-specs/specs/kilo/new-ironic-state-machine.html | 15:32 |
| *** rameshg87 is now known as rameshg87-brb | 15:32 | |
| jroll | rloo: oh, those. you should still be able to run node-validate at any time, though | 15:32 |
| rloo | jroll: are our existing validate() calls different than doing the VALIDAT* stuff in the new state machine | 15:32 |
| rloo | jroll: so after validating, passwords, etc, could still be changed. | 15:33 |
| jroll | rloo: AIUI, ENROLL -> MANAGEABLE requires validating the node | 15:33 |
| jroll | but you may want to validate later | 15:33 |
| jroll | for example, if you go back to MANAGEABLE and replace some hardware | 15:33 |
| rloo | jroll: but you can't go to VALIDAT* from MANAGEABLE. | 15:34 |
| rloo | jroll: so whatever happens, we'll still need to call the *validate()s before provisioning. | 15:34 |
| jroll | rloo: right, I don't think we should restrict the node-validate call to VALIDAT* states | 15:34 |
| jroll | rloo: for example, nova calls node-validate before deploying | 15:35 |
| gridinv | jroll: thanks, thought maybe something is going on nova side to make it easier. | 15:36 |
| jroll | gridinv: nope, no magic here :) | 15:36 |
| rloo | jroll: well, we need to distinguish what GET ../nodes/validate means, vs PUT ../states/provision/target=validate | 15:37 |
| jroll | rloo: in hindsight, I have no idea why VALIDAT* is a state | 15:37 |
| * jroll reads | 15:37 | |
| jroll | rloo: though I'm thinking that PUT /states/provision?target=manage from ENROLL just implicitly puts the node through these validat* states | 15:38 |
| rloo | jroll: we can always update the spec. But it does have an explicit 'validate' verb there. | 15:39 |
| jroll | rloo: ah | 15:39 |
| jroll | yeah, this seems weird to me | 15:39 |
| rloo | jroll: well, the idea could be that someone enrolls a bunch of nodes, and we want them to explicitly say 'ok, i have provided all the info you need and i think they are ready to be managed'. so yeah, maybe 'manage' might be better than 'validate'. | 15:40 |
| rloo | jroll: regardless, it is an explicit ack from the user that the nodes are ready to be managed. | 15:41 |
| rloo | jroll: right now, we kind of guess, do stuff or not, based on whether we have the required info. | 15:41 |
| jroll | rloo: indeed, though 'manage' is the verb to go from ABAILABLE -> MANAGE | 15:41 |
| rloo | jroll: you can use the same verb again, that isn't a problem. | 15:41 |
| rloo | jroll: eg, we use 'done' and 'fail'. | 15:42 |
| jroll | ah, true | 15:42 |
| *** rameshg87-brb is now known as rameshg87 | 15:43 | |
| rloo | jroll: it just seems murky if we can't 'guarantee' anything beyond VALIDATE. Eg, maybe we can't allow required driver_info properties to change after that, or only allow updates if the nodes are in eg 'enroll' state. dunno. | 15:44 |
| rloo | jroll: but i guess i digressed. i'm not sure if this helps at all wrt the problem rameshg87 was discussing. | 15:45 |
| jroll | rloo: dunno, should be able to change things like deploy ramdisk without deleting the node from the db (the only way to get to ENROLL is to create a new node) | 15:45 |
| * dtantsur is wondering when we are going to actually have ENROLL... | 15:46 | |
| rloo | jroll: well, we could change the code (we have the power!) to allow getting to ENROLL w/o deleting the node ;) | 15:46 |
| jroll | heh | 15:46 |
| jroll | rloo: I think that's the point of MANAGE. idk. | 15:46 |
| *** lazy_prince has quit IRC | 15:47 | |
| rloo | jroll: for me, that goes back to what VALIDATE does, and what it means (if anything) after a node has been VALIDATEd. | 15:47 |
| jroll | rloo: right | 15:47 |
| rloo | dtantsur: what does it mean to you, if we have nodes in ENROLL state? | 15:47 |
| rloo | dtantsur: like, we could add ENROLL now, and then transition them all to MANAGEABLE right away. | 15:48 |
| dtantsur | that confuses me as well actually | 15:48 |
| * jroll does not care for about this ENROLL state much, mostly because he doesn't care about inspection | 15:48 | |
| dtantsur | (and right now we're starting with available even and have to go one step back to be able to INSPECT) | 15:48 |
| dtantsur | jroll, ENROLL is not related to inspection at all | 15:49 |
| *** rameshg87 has quit IRC | 15:49 | |
| dtantsur | only discoverd can work with nodes without IPMI credentials (provided a user can press the physical button) | 15:49 |
| jroll | I thought the point of ENROLL was so you could throw nodes at ironic and figure out the details later | 15:49 |
| rloo | maybe people can hash out/revisit the states at the summit ;) | 15:49 |
| jroll | rloo: /me cries | 15:50 |
| rloo | jroll: ha ha. oh, sorry :-( | 15:50 |
| dtantsur | jroll, 1. create a node with driver only - ENROLL; 2. set IPMI credentials - MANAGEABLE; 3. inspect/add manually scheduling properties - AVAILABLE | 15:50 |
| dtantsur | that's how I see it ^^^ | 15:50 |
| jroll | dtantsur: ah. then I definitely don't care about ENROLL :P | 15:51 |
| rloo | dtantsur: so, 1. create a node with driver only - ENROLL; 2. set IPMI credentials in node.driver_info, issue 'validate' API, VALIDAT*->MANAGEABLE; 3. issue 'inspect' API, INSPECT*->MANAGEABLE; 4. issue 'provide' API, CLEAN*->AVAILABLE | 15:53 |
| dtantsur | makes sense | 15:53 |
| rloo | dtantsur: based on our state machine spec. | 15:53 |
| dtantsur | I'm a bit worried that we're likely to end up with incomplete version in K.. | 15:54 |
| rloo | dtantsur: the only thing about the above, is that discussion we had awhile ago, where you were going to not lock a node, and allow it to have different provision states I think, while the inspection was happening. | 15:54 |
| rloo | dtantsur: does it matter if all the states aren't implemented in K*? | 15:54 |
| *** hj-hp has joined #openstack-ironic | 15:55 | |
| jroll | gonna need a new spec for L | 15:55 |
| jroll | :P | 15:55 |
| * jroll runs away | 15:55 | |
| *** BadCub has quit IRC | 15:55 | |
| dtantsur | rloo, now the flow is: 1. create an empty node (AVAILABLE), 2. fill in IPMI creds (still AVAILABLE!), 3. issue MANAGE, 4. issue INSPECT, 5. issue PROVIDE | 15:56 |
| dtantsur | a bit ugly, isn't it? | 15:56 |
| rloo | dtantsur: oh | 15:56 |
| *** hj-hp has quit IRC | 15:56 | |
| victor_lowther | Good morning, Ironic | 15:56 |
| dtantsur | victor_lowther, morning | 15:57 |
| rloo | dtantsur: so it would be good to get the external API in K, so people don't have to change how they enroll etc their nodes in the future | 15:57 |
| NobodyCam | morning victor_lowther :) | 15:57 |
| rloo | morning victor_lowther | 15:57 |
| *** BadCub has joined #openstack-ironic | 15:58 | |
| *** hj-hp has joined #openstack-ironic | 16:01 | |
| *** ekarlso has quit IRC | 16:03 | |
| *** SpamapS has quit IRC | 16:03 | |
| *** SpamapS has joined #openstack-ironic | 16:04 | |
| *** ekarlso has joined #openstack-ironic | 16:07 | |
| *** ChuckC has joined #openstack-ironic | 16:08 | |
| jgrimm | is diskimage-builder ironic-agent valid way to build IPA deploy kernel/ramdisk? | 16:10 |
| jroll | it is valid, but it is huge (as in it requires 3gb of RAM to even boot) | 16:11 |
| jroll | I highly recommend using the coreos builder: https://github.com/openstack/ironic-python-agent/tree/master/imagebuild/coreos | 16:11 |
| jroll | jgrimm: ^ | 16:11 |
| jgrimm | jroll, ok.. yeah it seemed to work, but noticed imagebuild too | 16:11 |
| jroll | there's also a tarball available of the latest build from master if you aren't doing customizations | 16:11 |
| jgrimm | jroll, long term will both continue to work? | 16:11 |
| jgrimm | jroll, well, i'm looking to get this working over on our power 8 boxes, so sort of lacking the coreos/docker atm | 16:12 |
| jroll | jgrimm: long term, I know the coreos builder will continue to work, no idea about DIB | 16:12 |
| jroll | oh :( | 16:12 |
| jgrimm | jroll, heh.. honestly haven't looked to see how bad it will be.. but DIB seemed like an easy first target | 16:13 |
| jroll | jgrimm: ironic with power 8 is interesting though :D | 16:13 |
| *** ifarkas has quit IRC | 16:13 | |
| jgrimm | jroll, it has IPMI and netboot.. so really the image building is the only snag | 16:13 |
| *** JoshNang has joined #openstack-ironic | 16:14 | |
| devananda | dtantsur: initial state sould be ENROLLED ... | 16:16 |
| *** erwan_taf has joined #openstack-ironic | 16:16 | |
| dtantsur | yeah, it should be, but it's not right now IIRC | 16:16 |
| devananda | dtantsur: you are correct | 16:16 |
| devananda | dtantsur: we can't change the default state for older clients tho, so this goes back to the earlier conversation about versioning certain changes | 16:17 |
| dtantsur | right | 16:17 |
| devananda | eg, a client which passes the version >= X should get different behavior when enrolling nodes | 16:17 |
| devananda | so we need version support in the client to land to be able totest ENROLLED | 16:18 |
| dtantsur | so much to do... :( | 16:20 |
| *** ChuckC has quit IRC | 16:20 | |
| *** a1exhughe5 has joined #openstack-ironic | 16:20 | |
| *** athomas has quit IRC | 16:23 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Check UUID correctness for Glance images https://review.openstack.org/151951 | 16:24 |
| *** anderbubble has quit IRC | 16:29 | |
| *** anderbubble has joined #openstack-ironic | 16:30 | |
| *** spandhe has joined #openstack-ironic | 16:30 | |
| *** alexpilotti has quit IRC | 16:31 | |
| dtantsur | rloo, is it ok to approve https://review.openstack.org/#/c/159100/ with only one +2? | 16:32 |
| dtantsur | no, I"m not against, just checking :D | 16:32 |
| rloo | dtantsur: oh geez. my bad. | 16:34 |
| rloo | dtantsur: not sure what I was thinking. (Not thinking.) thx for pointing that out. | 16:34 |
| *** athomas has joined #openstack-ironic | 16:34 | |
| dtantsur | rloo, np) | 16:35 |
| *** spandhe has quit IRC | 16:35 | |
| *** david-lyle_afk has quit IRC | 16:37 | |
| *** Nisha has joined #openstack-ironic | 16:38 | |
| *** gridinv has quit IRC | 16:40 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Fix nits for supporting non-glance images https://review.openstack.org/160732 | 16:43 |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Add python-openstackclient plugin for ironic-discoverd https://review.openstack.org/157448 | 16:44 |
| *** rameshg87 has joined #openstack-ironic | 16:45 | |
| *** rameshg87 is now known as rameshg87-limite | 16:45 | |
| *** rameshg87-limite is now known as rameshg87 | 16:45 | |
| rameshg87 | JoshNang, you around ? | 16:45 |
| JoshNang | o/ | 16:45 |
| rameshg87 | hey | 16:45 |
| rameshg87 | JoshNang, can we talk about the zap thing ? | 16:45 |
| JoshNang | so i thought on how to boot the agent | 16:46 |
| rameshg87 | okay, and how is it ? | 16:46 |
| JoshNang | the agent driver will already be overriding execute_clean_step. if it detects the agent isn't booted, it'll boot the agent, and return that the command is going to be async. when the agent first checks in, it'll start executing that first step. | 16:46 |
| JoshNang | that won't power the agent off again at the end though | 16:47 |
| rameshg87 | JoshNang, so ironic would never know what all are the clean steps that agent executes ? | 16:48 |
| JoshNang | it will, right before it starts executing | 16:48 |
| openstackgerrit | Ramakrishnan G proposed openstack/ironic: Add driver interface for RAID configuration https://review.openstack.org/155230 | 16:48 |
| *** mtanino has joined #openstack-ironic | 16:49 | |
| rameshg87 | JoshNang, so what will get_clean_steps return ? | 16:49 |
| JoshNang | when the conductor does get_clean_steps, the agent driver will ask the agent for the steps and return those, and then those will be mixed in and save to node.driver_internal_info['clean_steps'] | 16:49 |
| rameshg87 | JoshNang, but conductor does get_clean_steps before executing the first step, right ? | 16:50 |
| * rameshg87 opens the cleaning review | 16:50 | |
| *** jistr has quit IRC | 16:50 | |
| JoshNang | correct | 16:50 |
| rameshg87 | JoshNang, so when are we actually booting the agent ? | 16:51 |
| JoshNang | ah i see what you're saying :P | 16:51 |
| *** killer_prince has joined #openstack-ironic | 16:51 | |
| *** killer_prince is now known as lazy_prince | 16:51 | |
| openstackgerrit | Dmitry Tantsur proposed stackforge/ironic-discoverd: Add python-openstackclient plugin for ironic-discoverd https://review.openstack.org/157448 | 16:51 |
| rameshg87 | JoshNang, i had one idea in mind but it conflicts with yours | 16:52 |
| rameshg87 | JoshNang, conflicts yours that agent should return the clean steps | 16:52 |
| JoshNang | well mine won't actually work :) so do tell | 16:52 |
| rameshg87 | JoshNang, we can define the agent zap/clean method in ironic within an interface (deploy, management) | 16:53 |
| rameshg87 | JoshNang, we decorate it with @boot_agent | 16:54 |
| *** jcoufal has quit IRC | 16:54 | |
| rameshg87 | JoshNang, that decorator will check if agent is booted and heartbeating | 16:54 |
| rameshg87 | JoshNang, if it is, it will call the method which will inturn send the command to the agent | 16:54 |
| rameshg87 | JoshNang, if it isn't, it will save the method name in driver internal info | 16:54 |
| rameshg87 | JoshNang, and return | 16:55 |
| JoshNang | that takes away a ton of the flexibility of inband cleaning/zapping. | 16:55 |
| rameshg87 | JoshNang, when agent boots and starts heartbeating, we call the method | 16:55 |
| *** erwan_taf has quit IRC | 16:55 | |
| trown | dtantsur: I am interested in the python-openstackclient, is there other client features to implement for discoverd outside of the patch 157448? | 16:55 |
| rameshg87 | JoshNang, yeah, that would require ironic to know (in code) what all clean/zap methods are available | 16:56 |
| dtantsur | trown, you may overtake 157448 if you want to. It lacks testing and API version support (e.g. http://docs.openstack.org/developer/python-openstackclient/plugins.html) | 16:56 |
| dtantsur | trown, if you don't want to, this patch is missing 2 arguments for setting IPMI credentials | 16:57 |
| JoshNang | right. i think the simpler answer is "power on the node before cleaning" | 16:57 |
| dtantsur | trown, these: https://github.com/stackforge/ironic-discoverd/blob/master/ironic_discoverd/client.py#L35 you can stack a patch on top of mine adding them | 16:57 |
| dtantsur | wdyt? | 16:57 |
| rameshg87 | JoshNang, yeah if node is powered on when cleaning starts, it's much simpler | 16:58 |
| rameshg87 | JoshNang, your idea will work straight away | 16:58 |
| trown | dtantsur: ya, I can do that | 16:58 |
| rameshg87 | JoshNang, why don't we check through all interfaces if some clean task is inband | 16:58 |
| rameshg87 | JoshNang, and if so boot the agent even before cleaning starts | 16:59 |
| trown | dtantsur: that being make a patch for the ipmi credential args | 16:59 |
| *** a1exhughe5 has quit IRC | 16:59 | |
| JoshNang | i think that's a good idea, but i'm going to go with the simpler "just power it on" for now, and loop back either at the end of the cycle/next cycle | 16:59 |
| dtantsur | trown, ack cool | 17:00 |
| rameshg87 | JoshNang, okay | 17:00 |
| rameshg87 | JoshNang, and for zapping ? | 17:00 |
| rameshg87 | JoshNang, because zapping is invoked from manageable state, right ? | 17:00 |
| rameshg87 | JoshNang, same thing ? just power on the node ? | 17:00 |
| JoshNang | correct, the same applies. | 17:00 |
| rameshg87 | JoshNang, okay makes sense for now | 17:01 |
| rameshg87 | JoshNang, we can better this in the next attempt | 17:01 |
| JoshNang | ++ | 17:01 |
| JoshNang | and thanks for pointing that out. i definitely would have missed it. i'm so used to the agent always running | 17:01 |
| dtantsur | see you tomorrow folks | 17:01 |
| rameshg87 | JoshNang, :) | 17:01 |
| rameshg87 | dtantsur, good night | 17:01 |
| dtantsur | trown, don't hesitate to make 157448 work while I'm out :) because now it does not | 17:02 |
| dtantsur | g'night! | 17:02 |
| *** dtantsur is now known as dtantsur|afk | 17:02 | |
| trown | dtantsur|afk: g'night | 17:02 |
| rameshg87 | JoshNang, so things look much easy now | 17:03 |
| *** andreykurilin_ has joined #openstack-ironic | 17:03 | |
| rameshg87 | JoshNang, so in do_node_clean() | 17:03 |
| rameshg87 | JoshNang, change the state to CLEANING and power on the node | 17:03 |
| rameshg87 | JoshNang, when agent starts heartbeating issue an rpc to start the cleaning again | 17:04 |
| rameshg87 | JoshNang, this time it will find node is powered on and agent is heartbeating | 17:04 |
| rameshg87 | JoshNang, and it will start with the clean steps | 17:04 |
| rameshg87 | JoshNang, makes sense ? | 17:04 |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Fix nits for supporting non-glance images https://review.openstack.org/160732 | 17:05 |
| JoshNang | rameshg87: correct. i'll have to add some thing to handle the DIB pxe disk | 17:05 |
| JoshNang | *pxe ramdisk | 17:05 |
| rameshg87 | JoshNang, but we wouldn't support inband operations by pxe ramdisk, right ? | 17:05 |
| JoshNang | i'm going to look at erasing the disk via iscsi | 17:06 |
| JoshNang | and it would never heartbeat back.. | 17:06 |
| rameshg87 | JoshNang, okay | 17:07 |
| rameshg87 | JoshNang, so i am ready to take up something if you want to start looking at something else | 17:07 |
| JayF | JoshNang: given iscsi is moving to the IPA ramdisk; should you even work on supporting cleaning in the bash ramdisk? | 17:08 |
| JoshNang | JayF: hmm that would be preferable | 17:08 |
| rameshg87 | JayF, +1, we might soon deprecate pxe ramdisk in a release, right ? | 17:10 |
| rameshg87 | i mean bash ramdisk | 17:10 |
| JoshNang | works for me! | 17:11 |
| *** alexpilotti has joined #openstack-ironic | 17:11 | |
| *** andreykurilin_ has quit IRC | 17:12 | |
| *** romcheg has quit IRC | 17:13 | |
| *** andreykurilin_ has joined #openstack-ironic | 17:13 | |
| *** russell_h has joined #openstack-ironic | 17:14 | |
| *** russell_h has quit IRC | 17:14 | |
| *** russell_h has joined #openstack-ironic | 17:14 | |
| *** jjohnson2 has quit IRC | 17:15 | |
| *** chlong has quit IRC | 17:17 | |
| rameshg87 | JoshNang, just let me know if you would like to offload any of ironic-zap related part to me :) | 17:17 |
| JoshNang | rameshg87: sure! let me see how far I get today (it's all I'm working on) | 17:17 |
| rameshg87 | JoshNang, my spec depends on ironic inband zapping, i would be more than happy to help | 17:17 |
| rameshg87 | JoshNang, great..thanks | 17:18 |
| *** Nisha_away has joined #openstack-ironic | 17:18 | |
| rameshg87 | Nisha, you around ? | 17:19 |
| *** Nisha has quit IRC | 17:19 | |
| rameshg87 | vdrok, https://review.openstack.org/#/c/160732/4/doc/source/deploy/install-guide.rst | 17:21 |
| rameshg87 | vdrok, doesn't it assume that we use single conductor for standalone setup ? | 17:21 |
| *** Nisha has joined #openstack-ironic | 17:22 | |
| *** Nisha_away has quit IRC | 17:23 | |
| vdrok | rameshg87, I haven't tested it, but I think it should work with multiple conductors | 17:23 |
| *** ChuckC has joined #openstack-ironic | 17:23 | |
| rameshg87 | vdrok, if there are multiple conductors, each node may be deployed in any of the conductor | 17:24 |
| rameshg87 | vdrok, so the tftp_server entry of dhcp option is not predictable and needs to be changed dynamically | 17:24 |
| rameshg87 | vdrok, right ? | 17:24 |
| vdrok | rameshg87, hm, yeah it seems so | 17:25 |
| *** rameshg87_ has joined #openstack-ironic | 17:27 | |
| *** hj-hp has quit IRC | 17:29 | |
| *** rameshg87 has quit IRC | 17:29 | |
| *** hj-hp has joined #openstack-ironic | 17:30 | |
| *** Nisha_away has joined #openstack-ironic | 17:30 | |
| *** Nisha has quit IRC | 17:32 | |
| *** vdrok is now known as vdrok_afk | 17:33 | |
| *** rameshg87_ is now known as rameshg87 | 17:34 | |
| rameshg87 | vdrok_afk, okay | 17:34 |
| vdrok_afk | rameshg87, so you think it should be mentioned there? | 17:34 |
| *** Nisha_away has quit IRC | 17:34 | |
| rameshg87 | vdrok_afk, yeah may be a note if you feel it's valid :) | 17:34 |
| *** Nisha_away has joined #openstack-ironic | 17:35 | |
| rameshg87 | vdrok_afk, beceause i haven't tried standalone + multiple conductors without neutron | 17:35 |
| *** anderbubble_ has joined #openstack-ironic | 17:36 | |
| vdrok_afk | vdrok_afk, yeah, me too, but it seems that youre correct, I always miss something :( | 17:36 |
| vdrok_afk | oops, rameshg87 ^ | 17:36 |
| *** anderbubble has quit IRC | 17:37 | |
| *** anderbubble_ is now known as anderbubble | 17:37 | |
| rameshg87 | vdrok_afk, :) | 17:37 |
| *** kkoski1 has quit IRC | 17:42 | |
| vdrok_afk | rameshg87, although I think it will work with multiple conductors if there are no conductors that enable same driver | 17:43 |
| rameshg87 | vdrok_afk, yeah in that case it might | 17:44 |
| rameshg87 | vdrok_afk, but then if 2 conductors don't run same driver - purpose of multiple conductor environment is defeated :) | 17:44 |
| rameshg87 | vdrok_afk, because nodes can't be taken over by another conductor | 17:44 |
| vdrok_afk | rameshg87, yup, but load can be reduced? | 17:45 |
| rameshg87 | vdrok_afk, yeah that's correct | 17:46 |
| *** romcheg has joined #openstack-ironic | 17:46 | |
| openstackgerrit | Vladyslav Drok proposed openstack/ironic: Fix nits for supporting non-glance images https://review.openstack.org/160732 | 17:52 |
| *** romcheg1 has joined #openstack-ironic | 17:52 | |
| *** romcheg has quit IRC | 17:52 | |
| vdrok_afk | rameshg87, as for your comment about manager.py here https://review.openstack.org/#/c/151951/ - invalid parameter value will be thrown on driver.deploy.validate step | 17:55 |
| vdrok_afk | rameshg87, and will be converted to InstanceDeployFailure | 17:56 |
| rameshg87 | vdrok_afk, that assumes that deploy driver does call is_glance_image | 17:56 |
| rameshg87 | vdrok_afk, but what if it doesn't ? | 17:56 |
| *** spandhe has joined #openstack-ironic | 17:56 | |
| rameshg87 | vdrok_afk, can we assume that ? :) | 17:57 |
| vdrok_afk | rameshg87, all the current drivers do, but i see, will fix it also, thanks | 17:57 |
| rameshg87 | vdrok_afk, idk, may be better get other's thoughts as well | 17:57 |
| * rameshg87 goes to sleep | 17:59 | |
| rameshg87 | vdrok_afk, i will check it tomorrow, just let me know whatever you decide .. | 18:00 |
| *** jlvillal_sc has joined #openstack-ironic | 18:00 | |
| rameshg87 | good night ironic | 18:00 |
| *** rameshg87 has quit IRC | 18:00 | |
| *** hj-hp has quit IRC | 18:01 | |
| NobodyCam | night rameshg87 | 18:01 |
| *** jlvillal_sc is now known as jlvillal_pdx | 18:03 | |
| *** jlvillal_pdx is now known as jlvillal_trainin | 18:04 | |
| *** Marga_ has joined #openstack-ironic | 18:07 | |
| *** Marga_ has quit IRC | 18:07 | |
| *** Marga_ has joined #openstack-ironic | 18:08 | |
| *** harlowja_away is now known as harlowja_ | 18:12 | |
| * jlvillal_trainin prepares for two days of training on OpenStack.... | 18:13 | |
| NobodyCam | jlvillal_trainin: cool | 18:13 |
| NobodyCam | and you have many nicks now | 18:13 |
| jlvillal_trainin | NobodyCam: I hope so. it is done by Mirantis. I can use the training! | 18:13 |
| NobodyCam | :) | 18:13 |
| jlvillal_trainin | NobodyCam: Yeah, I need to setup an IRC bouncer one of these days... | 18:14 |
| NobodyCam | :) | 18:14 |
| jlvillal_trainin | I heard jroll uses one. I wonder which one. | 18:14 |
| devananda | jlvillal_trainin: irssi + google notifier + screen | 18:14 |
| devananda | is what I use | 18:14 |
| NobodyCam | screen + cloud instance for me | 18:15 |
| jlvillal_trainin | devananda: Thanks. I will look into it. | 18:15 |
| jlvillal_trainin | NobodyCam: Thanks. | 18:15 |
| jroll | jlvillal_trainin: I use znc. also tmux + weechat | 18:15 |
| jlvillal_trainin | jroll: I'm already a tmux fan :) I've been using Chatzilla for years. But when Firefox crashes so does Chatzilla :( | 18:16 |
| *** alexpilotti_ has joined #openstack-ironic | 18:16 | |
| *** alexpilotti has quit IRC | 18:16 | |
| *** alexpilotti_ is now known as alexpilotti | 18:16 | |
| *** MattMan has quit IRC | 18:17 | |
| *** hj-hp has joined #openstack-ironic | 18:18 | |
| jroll | jlvillal_trainin: ya, I'm a huge fan of weechat | 18:18 |
| jlvillal_trainin | jroll: Thanks! | 18:19 |
| jroll | :) | 18:19 |
| *** andreykurilin_ has quit IRC | 18:22 | |
| *** andreykurilin_ has joined #openstack-ironic | 18:23 | |
| *** achanda has joined #openstack-ironic | 18:23 | |
| openstackgerrit | Chris Krelle proposed openstack/ironic: Check temp dir is writable https://review.openstack.org/160383 | 18:23 |
| *** athomas has quit IRC | 18:26 | |
| *** hj-hp has quit IRC | 18:41 | |
| *** ijw has joined #openstack-ironic | 18:44 | |
| openstackgerrit | Josh Gachnang proposed openstack/ironic: Fix nits in cleaning https://review.openstack.org/160599 | 18:48 |
| *** ijw has quit IRC | 18:48 | |
| *** Nisha_away has quit IRC | 18:49 | |
| adam_g | is the removal of 'rootfstype=ramfs' @ https://review.openstack.org/#/c/155728/16/ironic/drivers/modules/pxe_config.template required to boot the agent ramdisk on a machine with sufficient memory? | 18:54 |
| NobodyCam | adam_g: I've seen that in a patch | 18:54 |
| adam_g | it was removed from our default params @ 155728 | 18:55 |
| adam_g | noticed i cant boot my existing ramdisk anymore without re-adding it or bumping ram | 18:55 |
| *** achanda has quit IRC | 18:57 | |
| jroll | adam_g: I think lucas put it in dsg where needed | 18:57 |
| jroll | or something | 18:57 |
| *** achanda has joined #openstack-ironic | 18:57 | |
| adam_g | yeah, the VM memory got bumped back up to 1024 for one job | 18:58 |
| adam_g | but the job where it uses the default (512MB) is failing | 18:58 |
| adam_g | i just dont know if removal of that parameter is required to use the IPA ramdisk, or if we can leave it for both | 18:58 |
| jroll | I think it was iirc | 18:58 |
| jroll | which job uses the default? | 18:59 |
| adam_g | the parallel job, which im going to soon propose we drop... but more concerning is that the devstack defaults no longer work | 18:59 |
| adam_g | ill wait for lucas to get back and see what he thinks | 19:00 |
| *** lazy_prince has quit IRC | 19:00 | |
| jroll | adam_g: yeah, he's back tomorrow I think | 19:01 |
| *** gridinv has joined #openstack-ironic | 19:06 | |
| NobodyCam | brb | 19:08 |
| *** andreykurilin_ has quit IRC | 19:08 | |
| *** kkoski has joined #openstack-ironic | 19:12 | |
| *** hj-hp has joined #openstack-ironic | 19:14 | |
| *** hj-hp has quit IRC | 19:27 | |
| *** killer_prince has joined #openstack-ironic | 19:35 | |
| *** killer_prince is now known as lazy_prince | 19:35 | |
| *** lsmola has quit IRC | 19:36 | |
| *** dtantsur|afk has quit IRC | 19:36 | |
| *** ijw has joined #openstack-ironic | 19:38 | |
| *** dtantsur has joined #openstack-ironic | 19:41 | |
| *** ijw has quit IRC | 19:42 | |
| NobodyCam | brb again | 19:43 |
| *** ijw has joined #openstack-ironic | 19:45 | |
| *** lsmola has joined #openstack-ironic | 19:49 | |
| *** ijw has quit IRC | 19:49 | |
| *** romcheg1 has quit IRC | 19:51 | |
| *** Marga_ has quit IRC | 19:54 | |
| *** Marga_ has joined #openstack-ironic | 19:54 | |
| * NobodyCam is back | 20:03 | |
| *** ChuckC has quit IRC | 20:05 | |
| *** anderbubble has quit IRC | 20:10 | |
| NobodyCam | lintan: are you by chance around? | 20:11 |
| *** Marga_ has quit IRC | 20:15 | |
| *** Marga_ has joined #openstack-ironic | 20:16 | |
| *** erwan_taf has joined #openstack-ironic | 20:19 | |
| *** jlvillal_trainin has quit IRC | 20:24 | |
| *** jlvillal_pdx has joined #openstack-ironic | 20:29 | |
| *** achanda has quit IRC | 20:38 | |
| *** mrda-away is now known as mrda | 20:44 | |
| mrda | Morning Ironic | 20:44 |
| NobodyCam | morning mrda | 20:44 |
| mrda | hey NobodyCam | 20:44 |
| * NobodyCam 's house has sprung a leak. and it needs to be fixed. bbiab | 20:45 | |
| *** ijw has joined #openstack-ironic | 20:46 | |
| *** achanda has joined #openstack-ironic | 20:46 | |
| rloo | morning mrda | 20:49 |
| *** foexle has quit IRC | 20:50 | |
| *** ijw has quit IRC | 20:51 | |
| openstackgerrit | Josh Gachnang proposed openstack/ironic-python-agent: Fix incorrect IPA log message https://review.openstack.org/160977 | 20:55 |
| JoshNang | JayF: ^ | 20:55 |
| *** ijw has joined #openstack-ironic | 20:58 | |
| trown | do new features for the client library require specs? | 20:59 |
| trown | I was thinking of adding a plugin for the openstackclient to python-ironicclient | 21:00 |
| trown | like: https://github.com/stackforge/python-congressclient/tree/master/congressclient/osc | 21:00 |
| jroll | not usually | 21:00 |
| jroll | that seems fine to me, to not have a spec | 21:01 |
| trown | sweet | 21:01 |
| jroll | morning mrda :) | 21:03 |
| mrda | hey jroll | 21:05 |
| NobodyCam | wow gate really slow today :-p | 21:06 |
| *** gridinv has quit IRC | 21:07 | |
| *** gridinv has joined #openstack-ironic | 21:08 | |
| mrda | :( | 21:12 |
| mrda | How many days until release? :) | 21:12 |
| NobodyCam | :-p | 21:12 |
| *** wshao has joined #openstack-ironic | 21:25 | |
| *** ChuckC has joined #openstack-ironic | 21:26 | |
| *** wshao has quit IRC | 21:28 | |
| JayF | If I could get reviews on this simple change, I'd really appreciate it -> https://review.openstack.org/#/c/159986/ | 21:37 |
| JoshNang | JayF: sorry, thought I had already reviewed that :/ | 21:39 |
| JayF | it's alright | 21:39 |
| *** ndipanov has quit IRC | 21:39 | |
| JayF | I have been highly lax on upstream reviews recently, I deserve to wait | 21:39 |
| *** wshao has joined #openstack-ironic | 21:42 | |
| *** EmilienM has quit IRC | 21:44 | |
| *** EmilienM has joined #openstack-ironic | 21:44 | |
| *** dprince has quit IRC | 21:49 | |
| NobodyCam | lol nice 3 and a half hours to hit a gate failure... happy happy joy joy | 21:50 |
| jroll | JayF: +a | 21:50 |
| *** mtanino has quit IRC | 21:53 | |
| *** jgrimm is now known as zz_jgrimm | 21:57 | |
| NobodyCam | just looking at 145690, has anyone had to use --nic <net-id> when nova booting before? | 21:58 |
| *** wshao has quit IRC | 21:58 | |
| *** pelix has quit IRC | 21:58 | |
| *** mtanino has joined #openstack-ironic | 21:59 | |
| *** achanda has quit IRC | 22:01 | |
| *** andreykurilin_ has joined #openstack-ironic | 22:03 | |
| *** zz_jgrimm is now known as jgrimm | 22:03 | |
| *** achanda has joined #openstack-ironic | 22:07 | |
| NobodyCam | rloo: got a second for a quick question on a review? 159100 | 22:10 |
| rloo | NobodyCam: in a meeting but shoot | 22:11 |
| NobodyCam | i may have just answered my own question. but it was that I had a fear that the node_iter would affect a take over | 22:12 |
| rloo | NobodyCam: still have the fear? | 22:13 |
| *** devlaps has joined #openstack-ironic | 22:13 | |
| *** devlaps has quit IRC | 22:13 | |
| NobodyCam | less so now.. but still looking | 22:13 |
| NobodyCam | in to it | 22:13 |
| openstackgerrit | Josh Gachnang proposed openstack/ironic-python-agent: Add dispatch to all managers https://review.openstack.org/161001 | 22:14 |
| mrda | rloo: I agree with your comment about a spec being worthwhile for specifying microversion behaviour (ref 155624) | 22:17 |
| rloo | mrda: at the end of the day, we need to document whatever is decided. if folks are fine not having a spec to help with deciding, that is fine since it is so late in the game. just want to make sure we all understand/agree/whatever. | 22:18 |
| mrda | rloo: I'm concerned about 6 months time when we've all forgotten what we agreed on :) I think some documentation in this space is not a bad thought. | 22:19 |
| rloo | mrda: if that is wrt the python-ironicclient change, even nova's spec didn't help me understand how the client would be changed. | 22:19 |
| rloo | mrda: it is on my list todo, if no one else wants to document it ;) | 22:19 |
| mrda | rloo: Happy to assist / co-auth, or even just give it a thorough review | 22:21 |
| rloo | mrda: i'd be extremely happy if you auth-d it yourself ;) | 22:21 |
| mrda | lol | 22:21 |
| rloo | mrda: but if you don't want to, thx for the assist :-) | 22:21 |
| mrda | rloo: let's see if anything appears while you're sleeping overnight | 22:22 |
| mrda | rloo: so you think a cut-down version of this - https://github.com/openstack/nova-specs/blob/master/specs/kilo/approved/api-microversions.rst - is what is needed? Making sure we specify server/client behaviour? | 22:25 |
| *** wshao has joined #openstack-ironic | 22:25 | |
| rloo | mrda: I've seen two docs, that spec and the developer doc. sec. | 22:25 |
| rloo | mrda: docs.openstack.org/developer/nova/devref/api_microversions.html -- once we have things in place... | 22:26 |
| mrda | and there was also this: https://etherpad.openstack.org/p/kilo-nova-microversions | 22:27 |
| rloo | mrda: the ethernet pad, I didn't know about. thx I think ;) | 22:28 |
| mrda | I know the lead developer of the nova work - he's in the same city as me - so I'll catch up with him and discuss too | 22:29 |
| jroll | 21:58:16 NobodyCam | just looking at 145690, has anyone had to use --nic <net-id> when nova booting before? | 22:31 |
| jroll | ^ booting with admin in devsatck will do this | 22:31 |
| lifeless | jroll: if you are adin you can see all networks, so you'll get an error about nova not being able to guess otherwise | 22:31 |
| jroll | lifeless: right | 22:31 |
| openstackgerrit | Merged openstack/ironic-python-agent: Enable setting standalone mode via APARAMS https://review.openstack.org/159986 | 22:32 |
| *** anderbubble has joined #openstack-ironic | 22:32 | |
| mrda | rloo: Ok, I'll write something up today | 22:33 |
| *** wshao has quit IRC | 22:34 | |
| *** Marga_ has quit IRC | 22:35 | |
| devananda | so on the client change, i dont think the nova spec had much on it either? | 22:35 |
| rloo | devananda: yes, that's my understanding too. at least, i looked and I didn't see much there. | 22:36 |
| devananda | after chatting with a couple folks and thinking on the change naohirot proposed, I'm pretty strongly of hte opinion that | 22:38 |
| devananda | the client should, by default, specify EXACTLY the latest version it supports | 22:38 |
| mrda | devananda: I think so | 22:39 |
| *** kkoski has quit IRC | 22:39 | |
| devananda | my comment a week or so ago that we need to have "latest" support in the server is a convenient work around thta Nova defined. it helps developers, but I think it will hurt users if that's what the client passes in its request by default | 22:39 |
| devananda | i didn't realize it at the time, though | 22:39 |
| mrda | If we went down the whole semver thing for versioning, then we could handle this sort of mismatch better | 22:40 |
| mrda | i.e. we would talk about incompatible changes | 22:41 |
| *** kkoski has joined #openstack-ironic | 22:41 | |
| *** kkoski has quit IRC | 22:41 | |
| mrda | but the way things are, I think we need to have the client's maximum version being the ceiling, and the default, unless otherwise specified | 22:41 |
| *** kkoski has joined #openstack-ironic | 22:42 | |
| devananda | mrda: i dont see how semver would change that | 22:42 |
| *** anderbubble has quit IRC | 22:44 | |
| lifeless | agreed, 'latest' is for devs, not users | 22:44 |
| mrda | Well, right now, we just increment 1.X. We probably won't ever go 2.X. | 22:44 |
| devananda | lifeless: thus it is not a good client default | 22:45 |
| devananda | mrda: 2.x means we change the URI to /v2/ | 22:45 |
| mrda | Correct. What's the trigger for that? A backwards incompatible change? | 22:45 |
| devananda | correct | 22:45 |
| mrda | So if we removed, say, uuid as a (primary) field and indexed nodes on something else, then we would make that kind of change? | 22:46 |
| rloo | i think the client's default version (if the user doesn't specify) should be the same as the API default. Which is the minimum version. why do you think the client should default to the max version supported? | 22:46 |
| devananda | mrda: an older client must continue to be able to talk to the server, within the same API major version, if the server supports that major version | 22:46 |
| mrda | (not that we ever would so that in particular :) | 22:46 |
| lifeless | operationally doing a major API bump is super hard - see nova V3 right? | 22:46 |
| devananda | rloo: if the client supports newer functionality (eg, logical names) and so does the server, then-and-only-then it should be exposed, by default, to users | 22:47 |
| mrda | rloo: Because that would be the latest version, with the latest functionality, which is what everyone would want unless otherwise specified | 22:47 |
| mrda | lifeless: Agreed. I don't think we'll ever do it (happy to be proved wrong) | 22:47 |
| rloo | mrda, devananda: but given your logic, then the API request should default to the same max version too? | 22:47 |
| devananda | rloo: if the client defaults to passing the minimum version (1.1) then users will either: never see newer features, or: be required to specify exactly what version they want, which is bonkers. users should never have to know that | 22:47 |
| devananda | rloo: no. server API should default requests to min version, because we need t osupport older clients that did not pass any version | 22:48 |
| mrda | rloo: I think it should default to the latest version that both the client and server supports | 22:48 |
| rloo | devananda: so if a user uses the API directly...? | 22:48 |
| devananda | rloo: if they use curl, they can pass what ever header they want | 22:48 |
| devananda | rloo: if they write another client, ditto -- it's on them | 22:48 |
| rloo | devananda: I guess I don't see how users are different if they use the API vs the client. | 22:48 |
| devananda | rloo: they're not | 22:48 |
| rloo | devananda: if the user uses the client, they can pass whatever version they want | 22:48 |
| mrda | devananda: agreed. And if they don't specify anuything, they get 1.1 (aka Juno) | 22:49 |
| *** mjturek1 has quit IRC | 22:49 | |
| devananda | rloo: sure. they CAN. but we should not REQUIRE the user to know what precise version to send | 22:49 |
| rloo | devananda: but we REQUIRE the user to do that if they use the API. | 22:49 |
| devananda | the client actually supports some version greater than 1.1. it should request the server to provide that API | 22:49 |
| devananda | no more and no less | 22:49 |
| rloo | devananda: what am I missing? | 22:49 |
| devananda | rloo: no we dont | 22:49 |
| devananda | rloo: a client MAY specify a version header. in the absence of one, the server provides the minimum version supported | 22:50 |
| rloo | devananda: if the API request doesn't include the version header, they get the minimum. so the version needs to be specified. | 22:50 |
| mrda | rloo: they can specify 'latest' if they are using the REST API via curl, and hence not need to know the specific version | 22:50 |
| devananda | rloo: a user MAY override their client library's default, but in the absence of that override, the client library SHOULD inform the server of its greatest-supported version | 22:50 |
| *** openstackgerrit has quit IRC | 22:51 | |
| *** anderbubble has joined #openstack-ironic | 22:52 | |
| *** openstackgerrit has joined #openstack-ironic | 22:52 | |
| rloo | devananda: can't focus on this and the meeting i'm in. there is something I don't grok about the diff between a client (that issues an API request for the user) vs a user that issues an API request directly. | 22:52 |
| devananda | rloo: yea, i think it's in our terminology. I'm using "client (library)" to refer to python-ironicclient | 22:52 |
| devananda | rloo: and whether or not python-ironicclient should have a default, and what that default should be | 22:53 |
| *** andreykurilin_ has quit IRC | 22:55 | |
| mrda | So it looks like it might be useful to write this up. I'll put something together today. | 22:55 |
| devananda | mrda: ++ thanks much! | 22:55 |
| devananda | also this will be broadly useful as other projects start to implement it | 22:56 |
| mrda | np | 22:56 |
| devananda | i was thinking of doing a blog post on it, too | 22:56 |
| mrda | cool | 22:56 |
| devananda | but hten, i'm thinking of that a lot ... and not finding time ... | 22:56 |
| *** ijw has quit IRC | 22:57 | |
| *** ijw has joined #openstack-ironic | 23:00 | |
| * mrda wanders off to another meeting | 23:00 | |
| *** ijw_ has joined #openstack-ironic | 23:01 | |
| *** ijw has quit IRC | 23:05 | |
| *** spandhe has quit IRC | 23:05 | |
| rloo | meeting over, gotta go. will read mrda's write up. thx mrda! | 23:06 |
| NobodyCam | have a good night rloo | 23:07 |
| *** absubram has quit IRC | 23:08 | |
| NobodyCam | devananda: do we need this bug taged to k3? https://bugs.launchpad.net/ironic/+bug/1417939 | 23:16 |
| openstack | Launchpad bug 1417939 in Ironic "Switch to parallel periodic tasks once they're supported" [High,Triaged] | 23:16 |
| jroll | NobodyCam: depends on oslo | 23:17 |
| NobodyCam | I don't have good feeling that spec will land... ICBW | 23:18 |
| devananda | spec? | 23:18 |
| NobodyCam | that called out on the bug | 23:18 |
| devananda | the spec landed and was implemented in ironic | 23:18 |
| NobodyCam | not in oslo | 23:18 |
| devananda | oh | 23:18 |
| NobodyCam | thats why I asked about the bug targeted at k3 | 23:19 |
| devananda | dhellmann: any sense on https://review.openstack.org/#/c/134303/ -- spec in oslo, which we're depending on landing at some point | 23:19 |
| *** alexpilotti has quit IRC | 23:19 | |
| devananda | dhellmann: i believe we've landed a work-around in ironic for the lack of support for this in oslo, but it would be great to have an end in sight, even if thta's not kilo-3 | 23:20 |
| *** spandhe has joined #openstack-ironic | 23:24 | |
| *** jlvillal_ has joined #openstack-ironic | 23:30 | |
| *** kkoski has quit IRC | 23:30 | |
| *** chlong has joined #openstack-ironic | 23:31 | |
| *** anderbubble has quit IRC | 23:31 | |
| devananda | why do we have any RPC methods that are not asynchronous??? | 23:38 |
| NobodyCam | rloo: when your back I believe the -2 can be removed from: https://review.openstack.org/148804 | 23:42 |
| devananda | I'm looking at you, set-boot-device | 23:46 |
| *** achanda has quit IRC | 23:49 | |
| *** jgrimm is now known as zz_jgrimm | 23:50 | |
| rloo | NobodyCam: wrt 148804. Do we have a driver that does inspection? | 23:52 |
| devananda | also, wy does that return 204 ? | 23:53 |
| devananda | huh | 23:53 |
| devananda | well, that's awkwar | 23:53 |
| devananda | awkward | 23:53 |
| *** jlvillal_ has quit IRC | 23:54 | |
| NobodyCam | rloo: I was looking at https://review.openstack.org/#/q/topic:bp/ironic-node-properties-discovery,n,z | 23:54 |
| rloo | NobodyCam: I'll remove the -2 cuz I can't think now. I think that might need to be combined with the microversioning, client support of min/max versions, but not sure. | 23:55 |
| NobodyCam | rloo: please only remove if your comfortable doing so | 23:56 |
| rloo | NobodyCam: I removed it, but can always put it back later unless it gets merged ;) | 23:56 |
| NobodyCam | :) | 23:56 |
| *** achanda has joined #openstack-ironic | 23:56 | |
| NobodyCam | devananda: whats returning 204? | 23:57 |
| rloo | NobodyCam: gotta go out and shovel snow now, most likely won't be back til tomorrow. ciao. | 23:57 |
| *** rloo is now known as rloo_afk | 23:57 | |
| NobodyCam | have a good night rloo_afk :) | 23:57 |
| *** jlvillal_ has joined #openstack-ironic | 23:57 | |
| *** jlvillal_ has quit IRC | 23:57 | |
| *** jlvillal_ has joined #openstack-ironic | 23:58 | |
| *** jlvillal_ has quit IRC | 23:58 | |
| devananda | NobodyCam: set-node-boot-device | 23:58 |
| devananda | it's synchronous call to the BMC | 23:58 |
| devananda | the API returns status code 204 if the BMC completedthe request | 23:59 |
| devananda | we should NEVER have the APi blocking on a BMC call. EVAR | 23:59 |
| * devananda grumps | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!