| devananda | https://review.openstack.org/#/c/103164 | 00:00 |
|---|---|---|
| devananda | https://review.openstack.org/#/c/103165/ | 00:00 |
| devananda | https://review.openstack.org/#/c/103167/ | 00:00 |
| devananda | oh | 00:00 |
| devananda | i forgot to tag you | 00:00 |
| * devananda fails | 00:00 | |
| jroll | lol | 00:00 |
| devananda | wat! dansmith -1'd a patch with 0 lines of code change, lol | 00:01 |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Updated from global requirements https://review.openstack.org/106569 | 00:06 |
| devananda | mikal: also, this "propose it to nova to get reviews" is painful, just like we thought | 00:07 |
| devananda | mikal: because any review feedback has to be implemented twice | 00:08 |
| devananda | once on the next patchset of the review, and once in Ironic's tree | 00:08 |
| devananda | syncign these files is non-trivial because of changing import order | 00:09 |
| devananda | but now i'm just complaining :) | 00:09 |
| JayF | complaining is highly underrated | 00:09 |
| mikal | devananda: I am pretending to care as much as I can | 00:09 |
| mikal | Please hold | 00:09 |
| devananda | mikal: :) | 00:11 |
| mrda | devananda: please let me know what I can do to help. I'm happy to help be a patch monkey. | 00:15 |
| devananda | mrda: fantastic! | 00:15 |
| devananda | i'll summarize | 00:15 |
| devananda | mrda: white board lines 41 - 51 | 00:16 |
| devananda | "Code proposed" -- the patches to Nova | 00:16 |
| devananda | "Additional patches ..." -- the patches to Ironic which affect code that has been proposed to Nova (eg, the virt driver or scheduler bits) | 00:16 |
| devananda | any code proposed to ironic which touches a file under ironic/nova/* needs to be tracked | 00:17 |
| devananda | if it's approved, it needs to be merged into the propsals to nova | 00:17 |
| devananda | any feedback from the nova team that gets added to those nova reviews needs to be turned into a review in ironic, and fast-tracked | 00:17 |
| * mrda is reading the whiteboard now | 00:18 | |
| devananda | when the nova patches land, we freeze any outstanding patches in ironic, pivot tempest, and then delete the code from ironic | 00:18 |
| devananda | i'm hoping to do ^ that at the nova sprint | 00:18 |
| devananda | but we'll see | 00:18 |
| devananda | so this pain hopefully has an end in sight | 00:18 |
| devananda | mrda: so ... all taht is what i've been doing about once a week | 00:19 |
| devananda | i'll probably take one more stab at it next week, and then we have the summit | 00:19 |
| devananda | mrda: if we dont merge the driver at the sprint, we should think of a better long term solution -- and I'd be delighted if you take that on :) | 00:21 |
| mrda | "any feedback from the nova team that gets added to those nova reviews needs to be turned into a review in ironic, and fast-tracked" is the bit I'm unsure of. You mean create a new review in Ironic *each* time nova comments on {103164, 103165, 103167} ? | 00:21 |
| mrda | devananda: ^^ | 00:21 |
| devananda | mrda: also, your cache-rhe-client patch is a good example -- it's already on the whiteboard for tracking here | 00:21 |
| devananda | mrda: each time that nova's feedback on {...} results in a change to {...} -- yes | 00:21 |
| devananda | my goal is keep those patches {...} in sync with the current trunk of Ironic | 00:22 |
| mrda | devananda: could also include the ironicclient change change as a dependency, not requiring merging. but that's a moot point, because that should merge today (if you re-approve it to get around the intermittent devstack failure) - ref 104467 | 00:22 |
| mrda | :) | 00:22 |
| devananda | mrda: anyone can trigger a reverify | 00:23 |
| devananda | mrda: post a comment with "reverify bug ###" | 00:23 |
| devananda | with the appropriate bug # | 00:23 |
| mrda | oh, I thought reverify had gone away | 00:24 |
| devananda | it did. then infra brought it back .... | 00:24 |
| devananda | you can also use recheck still | 00:24 |
| devananda | that just makes it go through both sets of tests | 00:24 |
| mrda | devananda: do you have an example of nove feedback on the proposed patches resulting in an ironic review? | 00:24 |
| devananda | mrda: commit 0aec9f0a5305080115e128d0ec8fac9cf103b010 | 00:25 |
| mrda | I like to follow an approved example, if at all possible | 00:25 |
| devananda | and da967d77894be6f23d81fb5cc948f9d13898ba84 | 00:25 |
| devananda | mrda: so dansmith had feedback on the scheduler change (https://review.openstack.org/#/c/103165/) | 00:26 |
| devananda | mrda: that'd be a good place to start | 00:26 |
| mrda | ok, cool, I'll take a look and try and untangle | 00:27 |
| devananda | thanks! | 00:27 |
| *** harlowja_away is now known as harlowja | 00:30 | |
| openstackgerrit | Ellen Hui proposed a change to openstack/ironic-python-agent: Fix no IP on interface error https://review.openstack.org/107213 | 00:35 |
| *** Penick has joined #openstack-ironic | 00:35 | |
| openstackgerrit | Devananda van der Veen proposed a change to openstack/ironic: Import a few more fixes from the Nova driver https://review.openstack.org/107214 | 00:35 |
| devananda | mrda: another example ^ | 00:36 |
| mrda | devananda: thnx | 00:36 |
| devananda | I just applied dansmith's comments to the nova driver proposal, that's the "backport" if you will | 00:36 |
| devananda | heading out for dinner shortly | 00:38 |
| devananda | g'night all, see ya tmw! | 00:38 |
| mrda | night devananda | 00:39 |
| *** Penick has quit IRC | 00:48 | |
| openstackgerrit | A change was merged to openstack/ironic-python-agent: Fix no IP on interface error https://review.openstack.org/107213 | 00:52 |
| *** chuckC has quit IRC | 01:03 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Adding support for decommissioning https://review.openstack.org/104379 | 01:06 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Add versioning to Agent decommission https://review.openstack.org/106859 | 01:06 |
| *** Haomeng has quit IRC | 01:10 | |
| * mikal -2's an ironic nova driver patch | 01:33 | |
| mikal | devananda / mrda: so, if you're trying to keep the ironic version of the driver in sync, how do you detect when I upload a change to one of these reviews without telling you? | 01:38 |
| mrda | mikal: I'm just going to watch them, and do it handrollicly. | 01:39 |
| mikal | mrda: oh, that will be fun for you | 01:39 |
| mrda | mikal: We might come up with a different plan after mid-cycle, depending upon how things go by then. | 01:40 |
| *** nosnos has joined #openstack-ironic | 01:58 | |
| mikal | devananda: I don't suppose you're still around? | 02:02 |
| openstackgerrit | A change was merged to openstack/python-ironicclient: Expose auth_ref in ironicclient client object https://review.openstack.org/104467 | 02:12 |
| *** eghobo has quit IRC | 02:13 | |
| mrda | mikal: devananda has gone for the day, please try again tomorrow :) | 02:14 |
| *** chuckC has joined #openstack-ironic | 02:17 | |
| *** aswadr has joined #openstack-ironic | 02:19 | |
| *** haggybard has joined #openstack-ironic | 03:03 | |
| *** harlowja is now known as harlowja_away | 03:16 | |
| *** vinbs has joined #openstack-ironic | 03:24 | |
| *** ramineni has joined #openstack-ironic | 03:30 | |
| *** haggybard has quit IRC | 03:32 | |
| *** haggybard has joined #openstack-ironic | 03:32 | |
| *** nosnos has quit IRC | 03:43 | |
| *** bmahalakshmi has joined #openstack-ironic | 04:01 | |
| *** JayF has quit IRC | 04:02 | |
| *** Poornima has joined #openstack-ironic | 04:05 | |
| *** haggybard has quit IRC | 04:09 | |
| *** eghobo has joined #openstack-ironic | 04:09 | |
| *** eghobo has quit IRC | 04:14 | |
| *** eghobo has joined #openstack-ironic | 04:15 | |
| *** Nisha has joined #openstack-ironic | 04:15 | |
| *** Nisha has joined #openstack-ironic | 04:20 | |
| *** Nisha has quit IRC | 04:20 | |
| *** Nisha has joined #openstack-ironic | 04:21 | |
| *** eghobo has quit IRC | 04:27 | |
| *** killer_prince is now known as lazy_prince | 04:31 | |
| *** nosnos has joined #openstack-ironic | 04:32 | |
| *** rameshg87 has joined #openstack-ironic | 05:00 | |
| *** geekyogi has joined #openstack-ironic | 05:01 | |
| *** rakesh_hs has joined #openstack-ironic | 05:20 | |
| *** shausy has joined #openstack-ironic | 05:26 | |
| *** shausy2 has joined #openstack-ironic | 05:46 | |
| *** shausy has quit IRC | 05:47 | |
| *** chuckC has quit IRC | 05:51 | |
| *** chuckC has joined #openstack-ironic | 05:52 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Generic Hardware Discovery Bits https://review.openstack.org/102565 | 06:01 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 06:01 |
| *** bvivek has joined #openstack-ironic | 06:03 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 06:03 | |
| openstackgerrit | OpenStack Proposal Bot proposed a change to openstack/ironic: Imported Translations from Transifex https://review.openstack.org/106948 | 06:10 |
| *** geekyogi has quit IRC | 06:20 | |
| *** pcrews has quit IRC | 06:21 | |
| *** ndipanov_gone is now known as ndipanov | 06:34 | |
| *** Haomeng has joined #openstack-ironic | 06:38 | |
| *** jcoufal has joined #openstack-ironic | 06:38 | |
| openstackgerrit | Haomeng,Wang proposed a change to openstack/ironic: Implements send-data-to-ceilometer https://review.openstack.org/72538 | 06:39 |
| mrda | devananda: Just before I forget, I'm updating the whiteboard with the current Nova Driver progress. | 06:51 |
| *** d0ugal has quit IRC | 07:07 | |
| *** d0ugal has joined #openstack-ironic | 07:07 | |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 07:10 |
| *** malini1 has joined #openstack-ironic | 07:14 | |
| *** foexle has joined #openstack-ironic | 07:32 | |
| *** jistr has joined #openstack-ironic | 07:33 | |
| jcoufal | dtantsur|afk: hey, what was the issue yesterday? I am sorry I was burden in other issues | 07:39 |
| *** ChanServ changes topic to "Bare Metal Provisioning | Status: http://bit.ly/ironic-whiteboard | Docs: http://docs.openstack.org/developer/ironic/ | Bugs: https://bugs.launchpad.net/ironic" | 07:54 | |
| ifarkas | morning Ironic | 07:56 |
| mrda | Morning ifarkas | 07:57 |
| Haomeng | morning ifarkas mrda :) | 07:57 |
| ifarkas | morning mrda Haomeng :-) | 07:58 |
| mrda | Hi Haomeng! | 07:59 |
| *** romcheg1 has joined #openstack-ironic | 08:00 | |
| ifarkas | devananda, hi, re the drac power driver. the code is up in wip, it's only missing tests. but because I will be on vacation next week, I am more confident that it will land in J3 | 08:00 |
| *** geekyogi has joined #openstack-ironic | 08:00 | |
| mrda | Great stuff ifarkas! | 08:04 |
| *** igordcard has joined #openstack-ironic | 08:05 | |
| ifarkas | mrda, thanks! ;-) | 08:05 |
| *** dtantsur|afk is now known as dtantsur | 08:06 | |
| dtantsur | Morning Ironic! | 08:06 |
| mrda | hi dtantsur | 08:08 |
| dtantsur | mrda, hi :) | 08:09 |
| romcheg1 | Morning dtantsur and mrda! | 08:09 |
| mrda | hey romcheg1 | 08:09 |
| dtantsur | romcheg1, morning :) | 08:09 |
| * romcheg1 wonders where another IRC client is running | 08:09 | |
| *** derekh_ has joined #openstack-ironic | 08:10 | |
| mrda | ok, dinner time. I'll be back later. | 08:17 |
| *** mrda is now known as mrda-afk | 08:17 | |
| Haomeng | :) | 08:19 |
| Haomeng | ifarkas: :) | 08:19 |
| openstackgerrit | Haomeng,Wang proposed a change to openstack/ironic: Implements send-data-to-ceilometer https://review.openstack.org/72538 | 08:21 |
| *** Poornima has quit IRC | 08:26 | |
| *** lucasagomes has joined #openstack-ironic | 08:39 | |
| *** eglynn has joined #openstack-ironic | 08:40 | |
| eglynn | Haomeng: good morning sir! | 08:41 |
| Haomeng | eglynn: morning:) | 08:41 |
| eglynn | Haomeng: just a quick question on https://blueprints.launchpad.net/ironic/+spec/send-data-to-ceilometer | 08:41 |
| dtantsur | jcoufal, hi! The thing is that folks do not like my spec on discovery and while it's not the end of the world :) and won't change much from UI pov, we're in the beginning again | 08:41 |
| eglynn | Haomeng: it's targetted at juno-3, yet the work looks very well progressed | 08:41 |
| dtantsur | jcoufal, so it's going to take time, and I hope I won't learn after 3 more weeks that my suggestions sucks again :) | 08:41 |
| eglynn | Haomeng: might this be the first BP in openstack history to be bumped *forward* a milestone? ;) | 08:42 |
| Haomeng | eglynn: what is the due date for juno-3, I think almost code is done, I am preparing the ut code now | 08:42 |
| dtantsur | jcoufal, I'll try to show up in the office today, so we can talk about details, if you feel like | 08:42 |
| eglynn | Haomeng: cool :) ... https://wiki.openstack.org/wiki/Juno_Release_Schedule | 08:43 |
| eglynn | Haomeng: juno-2 == July 24th | 08:43 |
| eglynn | Haomeng: juno-3 == September 4th | 08:43 |
| jcoufal | dtantsur: alright | 08:43 |
| Haomeng | eglynn: thk | 08:43 |
| Haomeng | eglynn: this is my patch - https://review.openstack.org/#/c/72538/ | 08:43 |
| eglynn | Haomeng: so I guess if that patch happens to land before Jul 24th, we could call it a juno-2 feature? | 08:43 |
| *** romcheg2 has joined #openstack-ironic | 08:43 | |
| eglynn | Haomeng: ... reason I'm asking is that the corresponding ceilo BP is targetted at juno-2 | 08:44 |
| Haomeng | eglynn: ok, I will try my best to close it before Jul 24th:) | 08:44 |
| Haomeng | eglynn: need ut and code review | 08:44 |
| eglynn | Haomeng: excellent, thank you sir! | 08:44 |
| rameshg87 | good morning ironic | 08:44 |
| Haomeng | eglynn: almost code is done | 08:44 |
| *** romcheg1 has quit IRC | 08:44 | |
| Haomeng | eglynn: wel, that is my work:) | 08:44 |
| rameshg87 | had a question, can we add dependency for a review on 2 reviews ? :-) | 08:45 |
| eglynn | Haomeng: BTW cdent has been testing with your patch, has declared "white smoke" :) | 08:45 |
| dtantsur | rameshg87, morning, no you can't | 08:45 |
| rameshg87 | dtantsur: i wanted to raise a new review dependant on https://review.openstack.org/105413 and https://review.openstack.org/105795 | 08:45 |
| rameshg87 | dtantsur: but they themselves are independent | 08:45 |
| Haomeng | eglynn: that is because the code cdent used is old version, that is havana code, I have change it after my spec is approved, so now, the new patch is almost fine:) | 08:46 |
| dtantsur | rameshg87, the only way for you is to ask the authors to make them dependent. gerrit can't do better | 08:46 |
| Haomeng | eglynn: sorry, not sure cdent is using my havana old code:) | 08:46 |
| eglynn | Haomeng: yeah I think he's taking latest from gerrit and applying a few manual changes | 08:47 |
| Haomeng | eglynn: the new code is based on the new spec which is approved | 08:47 |
| eglynn | Haomeng: his notes are here https://tank.peermore.com/tanks/cdent-rhat/20140715 | 08:47 |
| rameshg87 | dtantsur: i am the author of all 3 | 08:47 |
| dtantsur | rameshg87, well, then you can handle it :) | 08:47 |
| rameshg87 | dtantsur: so if i have A and B independent, and C depends on A and B | 08:48 |
| rameshg87 | dtantsur: should I make B dependent on A ? | 08:49 |
| rameshg87 | dtantsur: and then C depends on B ? | 08:49 |
| dtantsur | A -> B -> C, right | 08:49 |
| Haomeng | eglynn: yes, I am writing new ut code now, so I think ceilometer can take the message as input to do the testing - http://paste.openstack.org/show/86675/, this is the message I ironic will send to ceilometer | 08:49 |
| *** martyntaylor has joined #openstack-ironic | 08:49 | |
| rameshg87 | dtantsur: yes | 08:49 |
| rameshg87 | dtantsur: eventhough B is not dependent on A :-) | 08:49 |
| Haomeng | eglynn: I added comments for cdent for the sample message which will be sent to ceilometer | 08:50 |
| dtantsur | rameshg87, IPA folks have even larger line of patches, where only the last depends on all previous :) | 08:50 |
| rameshg87 | dtantsur: okay :-) | 08:50 |
| *** overlayer has joined #openstack-ironic | 08:50 | |
| dtantsur | relocating to the office, brb | 08:51 |
| eglynn | Haomeng: excellent :) | 08:51 |
| Haomeng | eglynn: so far we have one opening issue with the message for the fields user_id and project_id which required by ceilometer | 08:52 |
| eglynn | Haomeng: yeah, so are these ever set to concrete values in Ironic view of a node? | 08:52 |
| eglynn | Haomeng: or does it depend on nova's view of the node as an instance? | 08:52 |
| Haomeng | eglynn: I have comments about this issue in both my patch and ceilometer patch - https://review.openstack.org/#/c/72538/ "Patch Set 14" | 08:52 |
| eglynn | Haomeng: a-ha, k, I'll check your comments on gerrit, thanks for the heads-up | 08:53 |
| Haomeng | eglynn: not sure if ceilometer is easy to get user_id and project_id these value, if not we have to wait our ironic nova driver merged to nova tree, and add new patch for ironic nova driver to retrieve these values from nova instance | 08:53 |
| eglynn | Haomeng: OK, that would work ... i.e. ceilo could tolerate these values being unset initially | 08:54 |
| Haomeng | eglynn: np, that is for our ironic nova driver prority, so dont want to modify nova ironic driver before it is approved by nova, that is important for ironic, hope understand:) | 08:54 |
| Haomeng | eglynn: ok, thank you | 08:55 |
| *** romcheg1 has joined #openstack-ironic | 08:55 | |
| eglynn | Haomeng: yeah, understood about getting the nova driver landed first and foremost | 08:55 |
| Haomeng | eglynn: we can fill it after our ironic nova driver is land into nova tree:) | 08:55 |
| Haomeng | eglynn: yes, thanks for understanding | 08:55 |
| eglynn | Haomeng: though I assume some Ironic nodes would not yet be known to nova, yet still reported on via the IPMI notifications? | 08:55 |
| *** romcheg2 has quit IRC | 08:56 | |
| eglynn | Haomeng: e.g. soon after you plug in a new rack in the DC? | 08:56 |
| Haomeng | eglynn: yes, only the nodes which booted by nova, has instance_uuid, that can be call by ironic to get ipmi sensor data | 08:57 |
| Haomeng | eglynn: so here, what is your concern | 08:57 |
| eglynn | Haomeng: a-ha OK, sounds like my concern is not a concern :) ... (if the node has already been "taken in charge" by nova before IPMI notifications are emitted) | 08:58 |
| Haomeng | eglynn: yes, it is managed by nova already | 09:00 |
| Haomeng | eglynn: so should be correct status and context | 09:01 |
| eglynn | Haomeng: cool, then all is well IIUC :) ... thanks for the confirmation! | 09:01 |
| Haomeng | eglynn: welcome | 09:01 |
| Haomeng | eglynn: I will be away for a while to go home, and online after 1 hour, you can find me after 1 hour:) | 09:02 |
| Haomeng | eglynn: if has more to be discussed:) | 09:02 |
| Haomeng | eglynn: :) | 09:02 |
| openstackgerrit | Nisha Agarwal proposed a change to openstack/ironic-specs: Discover node properties at node-create/node-update https://review.openstack.org/100951 | 09:02 |
| eglynn | Haomeng: cool, thanks ... I think I have enough detail now but I'll give cdent a heads-up in case he has further questions | 09:03 |
| *** bvivek has quit IRC | 09:05 | |
| rameshg87 | dtantsur: request you to take a look ilo deploy design spec when you get some time: https://review.openstack.org/97744 | 09:05 |
| *** bvivek has joined #openstack-ironic | 09:06 | |
| *** igordcard has quit IRC | 09:10 | |
| *** bvivek has quit IRC | 09:10 | |
| *** pelix has joined #openstack-ironic | 09:10 | |
| *** malini1 has quit IRC | 09:12 | |
| *** bvivek has joined #openstack-ironic | 09:12 | |
| *** cgoncalves has joined #openstack-ironic | 09:34 | |
| *** vinbs has quit IRC | 09:39 | |
| *** vinbs has joined #openstack-ironic | 09:42 | |
| openstackgerrit | A change was merged to openstack/ironic-specs: Add support for retry on NodeLocked exceptions https://review.openstack.org/103996 | 09:49 |
| lucasagomes | Shrews, ^ | 09:50 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Rename/update common/tftp.py to common/pxe_utils.py https://review.openstack.org/103595 | 10:00 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add iPXE Installation Guide documentation https://review.openstack.org/106809 | 10:00 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Factor out deploy info from PXE driver https://review.openstack.org/100735 | 10:00 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add methods to ipmitool driver https://review.openstack.org/100364 | 10:00 |
| openstackgerrit | Lucas Alvares Gomes proposed a change to openstack/ironic: Add iPXE support for Ironic https://review.openstack.org/99318 | 10:00 |
| lucasagomes | URGH! rebasing | 10:00 |
| lucasagomes | jroll, ^ >.< | 10:01 |
| lucasagomes | ok code's intact, just a new review was put up | 10:03 |
| *** athomas has quit IRC | 10:08 | |
| *** martyntaylor has left #openstack-ironic | 10:10 | |
| *** eglynn is now known as eglynn-commute | 10:18 | |
| *** Mikhail_D_wk has quit IRC | 10:19 | |
| *** Nisha has quit IRC | 10:19 | |
| *** lazy_prince is now known as killer_prince | 10:20 | |
| *** Mikhail_D_ltp has quit IRC | 10:22 | |
| *** eglynn-commute has quit IRC | 10:24 | |
| *** athomas has joined #openstack-ironic | 10:26 | |
| *** mrda-afk is now known as mrda | 10:29 | |
| *** killer_prince is now known as lazy_prince | 10:32 | |
| *** Mikhail_D_ltp has joined #openstack-ironic | 10:40 | |
| *** Alexei_987 has joined #openstack-ironic | 10:42 | |
| *** overlayer has quit IRC | 10:49 | |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add support for creating vfat disk images https://review.openstack.org/105413 | 10:52 |
| *** ramineni has quit IRC | 10:54 | |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Review fixes from Nova scheduler reviews https://review.openstack.org/107316 | 10:54 |
| *** kpavel has joined #openstack-ironic | 10:55 | |
| *** Mikhail_D_wk has joined #openstack-ironic | 10:59 | |
| dtantsur | mrda, when you're online, can we discuss your comment and my answer on https://review.openstack.org/#/c/73121/ ? | 11:00 |
| mrda | dtantsur: sure | 11:01 |
| * mrda looks | 11:01 | |
| mrda | We're talking about the comment on https://review.openstack.org/#/c/73121/18/ironic/db/sqlalchemy/api.py | 11:02 |
| mrda | correct? | 11:02 |
| mrda | So my concern is the same as your response, but in opposite :) I'm concerned that we haven't caught all possible exceptions out of SQLAlchemy either now, or in a future SQLAlchemy release | 11:06 |
| mrda | if we make a programming mistake with this approach, we leak information | 11:06 |
| mrda | But in my approach, if we make a programming mistake we still mask all sensitive information | 11:07 |
| mrda | But I also agree with your comment - we don't want to wrap exception we're not expecting | 11:07 |
| dtantsur | mrda, I guess, SQLAlchemy will always derive from their base exception | 11:08 |
| *** bmahalakshmi has quit IRC | 11:08 | |
| dtantsur | mrda, I'm seriously concerned with masking programming errors, not related to SQLA | 11:08 |
| mrda | Since no-one has jumped to my defence in the review, I'll comment that we've discussed this in IRC, and that you'll go ahead with your original approach | 11:09 |
| mrda | Is that ok by you dtantsur ? | 11:09 |
| dtantsur | mrda, yes, thank you :) | 11:09 |
| mrda | Of course if we get a security advisory from this code I will ahve my review comments to fall back on :-) | 11:10 |
| mrda | *much joking here* | 11:10 |
| mrda | dtantsur: You now have my +1 FWIW | 11:12 |
| dtantsur | :) | 11:14 |
| kpavel | Hi. I used devstack to work with ironic and check it's abilities and it worked fine. Nodes deployed to "baremetal" instances mocked by VMs...etc. Now i want to use same devstack setup to control real Physical machines. I got to the point when the physical machine finishes the deployment and becomes Active and on it's screen i can see the prompt. The problem is it's not answering ping. No... | 11:17 |
| kpavel | ...network to it. I think that my network configuration on the compute/controller is wrong and it remains configured to work with mocked virtual instances. Any help/reference for correct network config much appreciated :) | 11:17 |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add support for interacing with swift https://review.openstack.org/105795 | 11:18 |
| dtantsur | lucasagomes, https://etherpad.openstack.org/p/ironic-hw-discovery-2 is how it's shaping. Is it correct by now? | 11:19 |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add IloDriver and its IloPower module https://review.openstack.org/89500 | 11:26 |
| *** bvivek has quit IRC | 11:26 | |
| openstackgerrit | Haomeng,Wang proposed a change to openstack/ironic: Implements send-data-to-ceilometer https://review.openstack.org/72538 | 11:29 |
| openstackgerrit | Ramakrishnan G proposed a change to openstack/ironic: Add support for creating vfat disk images https://review.openstack.org/105413 | 11:29 |
| *** geekyogi has quit IRC | 11:30 | |
| lucasagomes | will take a look 1 sec | 11:32 |
| *** rameshg87 has left #openstack-ironic | 11:35 | |
| openstackgerrit | Haomeng,Wang proposed a change to openstack/ironic: Implements send-data-to-ceilometer https://review.openstack.org/72538 | 11:41 |
| mrda | devananda: in around 9 hours from now I'd like to discuss host_state_cls in https://review.openstack.org/#/c/103165/4/nova/scheduler/ironic_host_manager.py | 11:41 |
| mrda | night all | 11:41 |
| *** mrda is now known as mrda-away | 11:42 | |
| viktors|afk | lucasagomes: hi! | 11:44 |
| *** viktors|afk is now known as viktors | 11:44 | |
| openstackgerrit | Victor Sergeyev proposed a change to openstack/ironic: Use opportunistic approach for migration testing https://review.openstack.org/107053 | 11:49 |
| kpavel | Hi. I used devstack to work with ironic and check it's abilities and it worked fine. Nodes deployed to "baremetal" instances mocked by VMs...etc. Now i want to use same devstack setup to control real Physical machines. I got to the point when the physical machine finishes the deployment and becomes Active and on it's screen i can see the prompt. The problem is it's not answering ping. No... | 12:08 |
| kpavel | ...network to it. I think that my network configuration on the compute/controller is wrong and it remains configured to work with mocked virtual instances. Any help/reference for correct network config much appreciated :) | 12:08 |
| *** bvivek has joined #openstack-ironic | 12:08 | |
| *** jdob has joined #openstack-ironic | 12:09 | |
| Shrews | lucasagomes: awesome! thx | 12:12 |
| *** vinbs has quit IRC | 12:12 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic-specs: Generic Hardware Discovery Bits https://review.openstack.org/107344 | 12:17 |
| *** lucasagomes is now known as lucas-lunch | 12:21 | |
| dtantsur | new incarnation of discovery spec ^^^, not 100% finished yet | 12:21 |
| * romcheg1 takes a look | 12:24 | |
| *** Haomeng|2 has joined #openstack-ironic | 12:29 | |
| *** Haomeng has quit IRC | 12:30 | |
| *** tzumainn has joined #openstack-ironic | 12:36 | |
| tzumainn | dtantsur, ping | 12:36 |
| dtantsur | tzumainn, here | 12:36 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic-specs: Generic Hardware Discovery Bits https://review.openstack.org/107344 | 12:37 |
| tzumainn | dtantsur, heya - so regarding node auto-discovery - I hear that the mechanism underneath may change from your proposal; what about things like tagging nodes to distinguish auto-discovered nodes from other nodes? | 12:38 |
| dtantsur | tzumainn, see link right above your message, this is a new proposal. A new field will be introduced on a Node. | 12:38 |
| *** jbjohnso has joined #openstack-ironic | 12:44 | |
| *** jbjohnso_ has joined #openstack-ironic | 12:45 | |
| *** jbjohnso has quit IRC | 12:49 | |
| *** nosnos has quit IRC | 12:49 | |
| tzumainn | dtantsur, so is that part relatively uncontroversial? | 12:50 |
| dtantsur | tzumainn, I hope so :) | 12:50 |
| tzumainn | lol | 12:51 |
| tzumainn | dtantsur, so from an api perspective, IronicNodes.list() or whatever would return both auto-discovered nodes and "normal" nodes, and there would be a flag that allows me to filter out one or the other if desired? | 12:53 |
| dtantsur | tzumainn, new nodes will have: maintenance=True, newly_discovered=True | 12:54 |
| tzumainn | dtantsur, does the maintenance flag already exist? | 12:56 |
| dtantsur | tzumainn, yes | 12:57 |
| tzumainn | ah, cool - so the intent behind the maintenance flag is that a user has to "verify" the node somehow and then turn off the flag, correct? | 12:57 |
| jcoufal | tzumainn: yes | 12:58 |
| jcoufal | tzumainn: but there are 2 tags | 12:58 |
| tzumainn | so the workflow here is a) auto-discover nodes b) user verifies a node c) the user sets both tags to False ? | 12:58 |
| jcoufal | maintenance and newly_discovered | 12:58 |
| jcoufal | tzumainn: if he wants to get them to production, I believe yes | 12:58 |
| jcoufal | dtantsur: can you confirm? | 12:58 |
| tzumainn | although according to the spec, if you turn off the maintenance flag, newly_discovered is also set to False automatically | 12:58 |
| jcoufal | interesting | 12:59 |
| jcoufal | I guess it makes sort of sense | 12:59 |
| dtantsur | jcoufal, tzumainn, that's how I see it, yes. newly_discovered makes no sense with maintenance=False and will be also set to False, but you can unset both to be on a safe side | 13:00 |
| tzumainn | dtantsur, okay, and by default, Node.list() or whatever would return all nodes, including maintenance=True, unless you set some sort of flag in the call to filter them out? | 13:01 |
| tzumainn | jcoufal, so my question then is - what should we do with nodes that might be under maintenance, but which don't have the auto_discovered flag set to true? | 13:02 |
| jcoufal | tzumainn: they are registered, we shouw them under registered | 13:02 |
| jcoufal | tzumainn: for now | 13:03 |
| dtantsur | tzumainn, I'm not sure about Node.list, probably by default it does not return things with maintenance=True | 13:03 |
| *** romcheg2 has joined #openstack-ironic | 13:03 | |
| tzumainn | dtantsur, okay, that owuld make sense - thanks! | 13:03 |
| dtantsur | there is a parameter for it | 13:03 |
| jcoufal | dtantsur: it should | 13:03 |
| jcoufal | dtantsur: those nodes are already registered | 13:03 |
| jcoufal | why to exclude nodes in maintenance? | 13:03 |
| dtantsur | jcoufal, I just don't remember, how it works now :) | 13:04 |
| jcoufal | dtantsur: ok, np :) | 13:04 |
| *** romcheg1 has quit IRC | 13:04 | |
| tzumainn | jcoufal, so would the registered nodes have a new status field? | 13:04 |
| tzumainn | maintenance/non-maintenance? | 13:04 |
| jcoufal | tzumainn: they should | 13:05 |
| tzumainn | okay, should I just stick that in then? | 13:05 |
| jcoufal | tzumainn: go ahead | 13:05 |
| tzumainn | works for me, thanks! | 13:05 |
| *** Poornima has joined #openstack-ironic | 13:10 | |
| *** lucas-lunch is now known as lucasagomes | 13:12 | |
| *** dtantsur is now known as dtantsur|lunch | 13:12 | |
| *** romcheg1 has joined #openstack-ironic | 13:15 | |
| *** romcheg2 has quit IRC | 13:16 | |
| *** gestahlt has joined #openstack-ironic | 13:27 | |
| gestahlt | Hi | 13:27 |
| gestahlt | Do i need a pxe server for ironic? | 13:27 |
| *** NobodyCam has quit IRC | 13:29 | |
| *** BadCub_ has quit IRC | 13:30 | |
| *** dtantsur|lunch is now known as dtantsur | 13:31 | |
| dtantsur | gestahlt, hi! For using deploy via PXE you need a TFTP server, right | 13:31 |
| viktors | lucasagomes: around? | 13:36 |
| gestahlt | dtantsur: Yeah, i actually want to use ironic to manage a docker node.. since i have to do docker via heat | 13:36 |
| lucasagomes | viktors, hey yes, a bit afk tho | 13:36 |
| *** BadCub has joined #openstack-ironic | 13:36 | |
| viktors | lucasagomes: hi! I just want to show you patch https://review.openstack.org/#/c/107053/ (Use opportunistic approach for migration testing). Please ping me, when you'll have a time. | 13:37 |
| lucasagomes | viktors, nice, thanks for that! I poke u after I finish what I'm doing here | 13:39 |
| viktors | lucasagomes: sure, thanks | 13:39 |
| *** lazy_prince is now known as killer_prince | 13:40 | |
| *** rloo has joined #openstack-ironic | 13:41 | |
| *** NobodyCam has joined #openstack-ironic | 13:42 | |
| *** romcheg1 has left #openstack-ironic | 13:42 | |
| *** bvivek has quit IRC | 13:43 | |
| *** killer_prince is now known as lazy_prince | 13:43 | |
| *** kpavel has quit IRC | 13:46 | |
| *** jcoufal has quit IRC | 13:50 | |
| *** jcoufal has joined #openstack-ironic | 13:51 | |
| *** cdent has joined #openstack-ironic | 13:53 | |
| *** rloo has quit IRC | 13:53 | |
| *** rloo has joined #openstack-ironic | 13:54 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Add newly_discovered column to Node object https://review.openstack.org/107389 | 13:55 |
| NobodyCam | morning iRonic | 13:58 |
| dtantsur | NobodyCam, morning :) | 14:01 |
| *** jistr has quit IRC | 14:02 | |
| *** jistr has joined #openstack-ironic | 14:03 | |
| *** kpavel has joined #openstack-ironic | 14:04 | |
| *** lazy_prince is now known as killer_prince | 14:09 | |
| NobodyCam | morning dtantsur | 14:13 |
| dtantsur | folks, what steps are required for adding a new field to something (say Node)? Updating DB model, model in objects and creating a migration is not enough even for tests to pass :) | 14:14 |
| jroll | morning ironic :) | 14:20 |
| ifarkas | dtantsur, what test is failing for you? 107389 seems to be alright in the gate | 14:20 |
| jroll | dtantsur: tests should pass for that, especially considering that we're not currently testing migrations :/ | 14:21 |
| dtantsur | jroll, morning, strange, for me it looks like new field is not returned from DB | 14:21 |
| dtantsur | ifarkas, it's local | 14:21 |
| jroll | dtantsur: oh, you also have to run migrations :) | 14:21 |
| jroll | here's an example of a patch to add a field https://review.openstack.org/#/c/79466/ | 14:22 |
| dtantsur | jroll, well, migrations don't seem to work for SQLite and I still do not know how to make tests work with something else | 14:22 |
| jroll | dtantsur: to run them... ironic-dbsync --config-file etc/ironic/ironic.conf.local | 14:23 |
| dtantsur | jroll, yeah, I'm already using your patch as a reference :) | 14:23 |
| dtantsur | lemme try... | 14:23 |
| jroll | hmm | 14:23 |
| jroll | I guess for sqlite you would nuke the db and run that | 14:23 |
| dtantsur | CRITI [ironic] NotImplementedError: No support for ALTER of constraints in SQLite dialec | 14:23 |
| dtantsur | no good, "upgrade 31baaf680d2b -> 3bea56f25597, add unique constraint to instance_uuid" does not work | 14:25 |
| jroll | after deleting the db, even? | 14:25 |
| dtantsur | jroll, ^^^ | 14:25 |
| dtantsur | yes | 14:25 |
| jroll | O.o | 14:25 |
| jroll | that's not scary or anything | 14:25 |
| NobodyCam | morning jroll | 14:26 |
| dtantsur | I guess I'm going to add try...catch NotImplementedError... to a failing migration | 14:27 |
| jroll | heya NobodyCam | 14:27 |
| jroll | dtantsur: :| | 14:27 |
| jroll | I would think that dbsync without a db would just do it all in one shot | 14:27 |
| jroll | and not go through migrations | 14:27 |
| jroll | but, who knows | 14:28 |
| *** jistr has quit IRC | 14:32 | |
| viktors | devananda: Hello! Are you around? | 14:34 |
| *** jcoufal has quit IRC | 14:35 | |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Fix database migration with SQLite https://review.openstack.org/107407 | 14:35 |
| dtantsur | jroll, well ^^^ | 14:35 |
| viktors | dtantsur: hi! Sorry, I missed the discussion, but why do you need migrations on SQLite? | 14:36 |
| openstackgerrit | Dmitry Tantsur proposed a change to openstack/ironic: Add newly_discovered column to Node object https://review.openstack.org/107389 | 14:36 |
| openstackgerrit | Syed Ismail Faizan Barmawer proposed a change to openstack/ironic-specs: UEFI support for Ironic deploy drivers https://review.openstack.org/99850 | 14:36 |
| dtantsur | viktors, to run tests. Anyway, if we support SQLite, it will be nice to have it working | 14:37 |
| *** cdent_ has joined #openstack-ironic | 14:38 | |
| viktors | dtantsur: as for test - what do you think, if we will generate db schema for tests on sqlite from model description? | 14:38 |
| *** shausy2 has quit IRC | 14:39 | |
| viktors | dtantsur: also I was pretty sure, that we don't use sqlite in production, or I miss something? | 14:39 |
| dtantsur | viktors, I would like it, unless we want to test migrations as well (which probably should be done separately) | 14:39 |
| dtantsur | viktors, of course not in prod, but can be helpful for playing with services locally as well | 14:39 |
| dtantsur | I suspec our quickstart guide may even recommend it | 14:39 |
| *** faizan has joined #openstack-ironic | 14:40 | |
| viktors | dtantsur: afaik, at the moment we test migrations on production databases (MySQL, PostgreSQL) | 14:40 |
| *** cdent has quit IRC | 14:40 | |
| *** cdent_ is now known as cdent | 14:40 | |
| viktors | dtantsur: sorry, one more time about quickstart guide ? | 14:41 |
| *** jbjohnso__ has joined #openstack-ironic | 14:41 | |
| *** jbjohnso_ has quit IRC | 14:45 | |
| viktors | dtantsur: also one more note about migrations on SQLite - Ironic use Alembic as migration tool, and (AFAIK) Alembic has very limited sqlite support | 14:45 |
| viktors | dtantsur: see for example https://bitbucket.org/zzzeek/alembic/issue/21/column-renames-not-supported-on-sqlite | 14:46 |
| *** pquerna has quit IRC | 14:46 | |
| dtantsur | viktors, http://docs.openstack.org/developer/ironic/dev/dev-quickstart.html#exercising-the-services-locally | 14:47 |
| dtantsur | it suggests mysql only as an option, defaulting to sqlite | 14:47 |
| *** pquerna has joined #openstack-ironic | 14:47 | |
| viktors | dtantsur: well, what do you think, if we will generate db schema from model description during the installation? It should be faster at least and suitable for sqlite. | 14:48 |
| dtantsur | viktors, +many to this :) | 14:49 |
| viktors | dtantsur: :) | 14:49 |
| dtantsur | lucasagomes, still waiting for your vote on https://review.openstack.org/#/c/107344/ ;) | 14:50 |
| *** JayF has joined #openstack-ironic | 14:51 | |
| *** jgrimm has joined #openstack-ironic | 14:51 | |
| viktors | dtantsur: we have a test case in oslo.db, which check, that DB schema, which was generated from migrations and from models are equal, so it should be totally safe to use a such approach. See test case at https://github.com/openstack/oslo.db/blob/master/oslo/db/sqlalchemy/test_migrations.py#L285 | 14:52 |
| dtantsur | viktors, wow cool! | 14:53 |
| NobodyCam | dtantsur: :-p Create new method in `ManagementInterface` ... https://github.com/openstack/ironic/blob/master/ironic/drivers/base.py#L357 | 14:53 |
| lucasagomes | dtantsur, I'm trying to put it in a ramdisk and see if we are not missing any gaps | 14:53 |
| NobodyCam | nm | 14:54 |
| NobodyCam | I not yet awake | 14:54 |
| jroll | lucasagomes, dtantsur, gah, I had thought of a reason to use dtantsur's method yesterday... now I forgot :/ | 14:55 |
| jroll | one quick thing, though, is passing back extra data to the node (like for the agent we pass back heartbeat_timeout) | 14:56 |
| lucasagomes | jroll, extra data? | 14:58 |
| *** Poornima has quit IRC | 14:58 | |
| lucasagomes | jroll, as part of the discovery? | 14:58 |
| jroll | lucasagomes: in the agent... we call vendor_passthru/lookup to find which node it is... the response is something like {'node': 'uuid-here', 'heartbeat_timeout': 300} | 14:58 |
| jroll | so that ironic can define the heartbeat timeout | 14:59 |
| jroll | that could be in dhcp configs but mehhhhhhhhhhh | 14:59 |
| *** geekyogi has joined #openstack-ironic | 14:59 | |
| lucasagomes | jroll, heh right, but to find which node it is we kinda agreed at looking at ports no? | 14:59 |
| dtantsur | well, my method was way more generic, though somewhat suboptimal | 14:59 |
| lucasagomes | also, if the machine is already registered | 15:00 |
| jroll | lucasagomes: right. I'm saying that would be a reason to put the logic on the ironic side | 15:00 |
| *** geekyogi has quit IRC | 15:00 | |
| lucasagomes | it's very unlikely that it will boot on the discovery ramdisk | 15:00 |
| lucasagomes | jroll, ah right yea | 15:00 |
| dtantsur | I raised this in review as well. Also, jroll, your review still suggests storing inventory. | 15:00 |
| jroll | lucasagomes: also, this is only tangentially related to discovery :) | 15:00 |
| jroll | dtantsur: I know it does, been focusing on internal things this week :/ | 15:00 |
| jroll | dtantsur: I'll update it soon | 15:01 |
| dtantsur | ack | 15:01 |
| NobodyCam | dtantsur: question on 107344, who Is going to press the power button to boot a new (undiscovered) node?? | 15:01 |
| jroll | likely after you're gone for the night, unfortunately | 15:01 |
| JayF | jroll: fyi; bouncer host server died last night, I'm trying to get back into internal irc :x | 15:01 |
| jroll | JayF: lol :( | 15:01 |
| openstackgerrit | Imre Farkas proposed a change to openstack/ironic-specs: DRAC management driver https://review.openstack.org/107033 | 15:01 |
| dtantsur | NobodyCam, alive user. I don't see other chances for now. | 15:01 |
| jroll | NobodyCam: buy better hardware that boots as soon as it's plugged in :) | 15:02 |
| NobodyCam | I was thinking arp ping the ipmi address range and get every reply for a unknowen mac, if the mac is unknown then use cypher zero to power on the node :-p | 15:03 |
| NobodyCam | but that may be too much :-p | 15:03 |
| JayF | NobodyCam: OpenCompute spec hardware is required to power-on and PXE boot all interfaces | 15:04 |
| JayF | NobodyCam: so there is a whole set of hardware that is just plug-in and it starts pxe | 15:04 |
| jroll | NobodyCam: lol | 15:04 |
| *** JayF has quit IRC | 15:04 | |
| *** ndipanov_ has joined #openstack-ironic | 15:05 | |
| faizan | morning NobodyCam, dtantsur | 15:08 |
| *** ndipanov has quit IRC | 15:08 | |
| *** athomas has quit IRC | 15:08 | |
| dtantsur | faizan, morning! | 15:08 |
| *** Poornima has joined #openstack-ironic | 15:08 | |
| dtantsur | NobodyCam, LOL | 15:08 |
| NobodyCam | morning faizan | 15:09 |
| NobodyCam | brb | 15:09 |
| faizan | dtantsur: NobodyCam: I have submitted the uefi support design spec after addressing devananda's review comments. https://review.openstack.org/99850 | 15:09 |
| faizan | kindly request you folks to take a look at this spec and give your comments. | 15:10 |
| *** gestahlt has quit IRC | 15:14 | |
| *** faizan has quit IRC | 15:14 | |
| *** ifarkas has quit IRC | 15:15 | |
| *** amitpp has joined #openstack-ironic | 15:16 | |
| *** mdorman has joined #openstack-ironic | 15:16 | |
| *** JayF has joined #openstack-ironic | 15:19 | |
| *** jistr has joined #openstack-ironic | 15:21 | |
| *** viktors is now known as viktors|afk | 15:21 | |
| *** JayF has quit IRC | 15:27 | |
| *** foexle has quit IRC | 15:29 | |
| *** ndipanov_ has quit IRC | 15:31 | |
| *** JayF has joined #openstack-ironic | 15:31 | |
| *** pcrews has joined #openstack-ironic | 15:32 | |
| *** pcrews has quit IRC | 15:38 | |
| *** Mikhail_D_ltp has quit IRC | 15:40 | |
| *** rakesh_hs has quit IRC | 15:42 | |
| *** pcrews has joined #openstack-ironic | 15:42 | |
| *** ndipanov_ has joined #openstack-ironic | 15:42 | |
| dtantsur | going now, see you :) | 15:44 |
| *** dtantsur is now known as dtantsur|afk | 15:44 | |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: Fix tear_down a node with missing info https://review.openstack.org/103685 | 15:47 |
| *** ndipanov_ has quit IRC | 15:56 | |
| *** ndipanov has joined #openstack-ironic | 15:58 | |
| *** ndipanov has quit IRC | 16:00 | |
| * NobodyCam now remembers why he dislikes windows so much | 16:00 | |
| *** matty_dubs is now known as matty_dubs|lunch | 16:01 | |
| NobodyCam | brb | 16:03 |
| *** Poornima has quit IRC | 16:08 | |
| *** chuckC has quit IRC | 16:10 | |
| *** malini1 has joined #openstack-ironic | 16:15 | |
| NobodyCam | grr... | 16:16 |
| Shrews | NobodyCam is either turning into a bear, or still battling windows | 16:21 |
| NobodyCam | it is hte latter :-p | 16:21 |
| NobodyCam | the* | 16:21 |
| *** derekh_ has quit IRC | 16:27 | |
| *** malini1 has quit IRC | 16:29 | |
| *** rakesh_hs has joined #openstack-ironic | 16:29 | |
| *** jistr has quit IRC | 16:30 | |
| *** bvivek has joined #openstack-ironic | 16:42 | |
| *** martyntaylor has joined #openstack-ironic | 16:46 | |
| devananda | mornign, all | 16:46 |
| *** Mikhail_D_ltp has joined #openstack-ironic | 16:46 | |
| *** foexle has joined #openstack-ironic | 16:47 | |
| NobodyCam | morning devananda :) | 16:47 |
| *** malini1 has joined #openstack-ironic | 16:48 | |
| *** Alexei_987 has quit IRC | 16:48 | |
| devananda | viktors|afk: hi! | 16:52 |
| devananda | 13:36:21 < gestahlt> dtantsur: Yeah, i actually want to use ironic to manage a docker node.. since i have to do docker via heat | 16:53 |
| devananda | anyone know who that ^ was? | 16:53 |
| NobodyCam | gah missed dtantsur|afk | 16:55 |
| devananda | kpavel: hi! did you have any success / response to your network question? | 16:55 |
| devananda | mrda-away: ping me when you're around. Yes, I'd like to talk about ironic_host_manager :) | 16:59 |
| Shrews | morning devananda | 17:01 |
| devananda | mornin, Shrews ! | 17:01 |
| *** matty_dubs|lunch is now known as matty_dubs | 17:02 | |
| NobodyCam | morning Shrews | 17:03 |
| *** cdent_ has joined #openstack-ironic | 17:06 | |
| devananda | Shrews: any thoughts on how we should test this functionality? https://review.openstack.org/#/c/105802/ | 17:07 |
| devananda | Shrews: i'd like to know that it doesn't break in the future | 17:07 |
| devananda | Shrews: it's an interaction between ironic and nova. clearly outside of unit tests, but falls into the grey arae sdague has been pushing out of tempest lately | 17:08 |
| *** cdent_ has quit IRC | 17:08 | |
| Shrews | devananda: hrm, i was pondering over that earlier this morning, actually. didn't come to any decision though. i was trying to figure out how'd that would even be used. we don't set node capabilities yet, do we? | 17:08 |
| *** foexle has quit IRC | 17:09 | |
| Shrews | or my grep-foo failed | 17:09 |
| devananda | Shrews: we dont, and we wont | 17:09 |
| devananda | it's there for deployers | 17:09 |
| Shrews | ah | 17:09 |
| *** cdent has quit IRC | 17:10 | |
| devananda | I call "ironic node-update $UUID add properties/capabilities="foo:bar,baz:kazoo" and this should work | 17:10 |
| Shrews | right. makes sense now | 17:10 |
| devananda | && I create a nova flavor with capabilities foo:bar | 17:10 |
| devananda | and then nova scheduler finds the right node | 17:10 |
| Shrews | devananda: without tempest... hrm... | 17:11 |
| devananda | so it's an integration test between ironic and nova-compute and nova-scheduler | 17:11 |
| *** cdent has joined #openstack-ironic | 17:11 | |
| devananda | git grep extra_specs | grep -c compute | 17:13 |
| devananda | 52 | 17:13 |
| devananda | git grep capabilit | grep -c compute | 17:13 |
| devananda | 0 | 17:13 |
| devananda | :( | 17:14 |
| Shrews | devananda: ironic seems to try harder than the other projects to "test all the things" | 17:14 |
| Shrews | which is good | 17:14 |
| devananda | Shrews: it's because we care | 17:14 |
| jroll | heh | 17:15 |
| Shrews | and you are our overlord | 17:15 |
| kpavel | devananda: Hi! No, i didn't have response yet :( Still straggling. | 17:15 |
| Shrews | i mean.... PTL | 17:15 |
| Shrews | :) | 17:15 |
| jroll | openstack has no concept of integration tests, IMO :( | 17:15 |
| Shrews | devananda: i don't have a good idea yet. perhaps something to add to the list next week? | 17:15 |
| devananda | jroll: read the recent ML thread about it? | 17:15 |
| jroll | e.g. when I add the configdrive feature... people are of the opinion that tempest shouldn't change. wat | 17:16 |
| devananda | jroll: right | 17:16 |
| jroll | devananda: no... will look | 17:16 |
| jroll | later | 17:16 |
| jroll | afk for a bit | 17:16 |
| devananda | jroll: I expect the same would be said about the support for CCF that I just added to nova.virt.ironic | 17:16 |
| jroll | also, you should ask dwalleck how he feels about tempest some time :) | 17:16 |
| *** openstackgerrit has quit IRC | 17:16 | |
| jroll | devananda: indeed | 17:16 |
| devananda | actually, me too. got a few things to do i nthe office this morning | 17:17 |
| NobodyCam | :) | 17:17 |
| Shrews | "office"? what is this strange word you bring before us? | 17:17 |
| devananda | kpavel: ok, i haven't done that myself, but in principle, i think it's just a matter of changes to neutron (simple, right? :) ) | 17:17 |
| *** openstackgerrit has joined #openstack-ironic | 17:17 | |
| kpavel | devananda: right... :) the big question s what change | 17:18 |
| *** harlowja_away is now known as harlowja | 17:18 | |
| NobodyCam | :( so many red X's on https://review.openstack.org/#/c/103164 | 17:18 |
| devananda | anyone gotten devstack (ironic, neutron, etc) to work booting phys hardware lately? I think we need some docs on that | 17:19 |
| *** harlowja is now known as harlowja_away | 17:19 | |
| *** rameshg87 has joined #openstack-ironic | 17:19 | |
| *** harlowja_away is now known as harlowja | 17:19 | |
| devananda | last time I did that in my lab was with tripleo, fwiw | 17:19 |
| * devananda needs to rebuild his lab after moving last week | 17:19 | |
| JayF | Yeah, that's something we wanna do too, for our hardware lab area | 17:19 |
| devananda | Shrews: OSSG meetup in seattle. and I need to print an expense report. | 17:20 |
| devananda | bbiab | 17:20 |
| *** hemna__ is now known as hemna | 17:25 | |
| rameshg87 | devananda: request you to take a look at ilo deploy spec https://review.openstack.org/97744 | 17:25 |
| NobodyCam | brb | 17:26 |
| *** rameshg87 has left #openstack-ironic | 17:30 | |
| *** lucasagomes is now known as lucas-dinner | 17:32 | |
| *** viktors has joined #openstack-ironic | 17:33 | |
| *** cgoncalves has left #openstack-ironic | 17:35 | |
| kpavel | devananda: Maybe if you have some reference to network configuration for physical i will manage, not only for devstack setup... | 17:38 |
| *** shakamunyi has joined #openstack-ironic | 17:39 | |
| *** shakamunyi has quit IRC | 17:39 | |
| *** shakamunyi has joined #openstack-ironic | 17:40 | |
| *** shakamunyi has quit IRC | 17:42 | |
| *** ifarkas has joined #openstack-ironic | 17:42 | |
| *** cdent_ has joined #openstack-ironic | 17:44 | |
| NobodyCam | lucas-dinner: are you out for the day | 17:44 |
| *** cdent has quit IRC | 17:45 | |
| *** cdent_ is now known as cdent | 17:45 | |
| *** amitpp has quit IRC | 17:47 | |
| Shrews | afk for a bit | 17:52 |
| *** Penick has joined #openstack-ironic | 17:55 | |
| *** cdent has quit IRC | 17:56 | |
| *** Penick has quit IRC | 17:59 | |
| *** Penick has joined #openstack-ironic | 18:02 | |
| ifarkas | devananda, hi, are you around? | 18:03 |
| NobodyCam | ifarkas: I think he's on his way into the office | 18:08 |
| ifarkas | NobodyCam, cool, thanks | 18:09 |
| devananda | back | 18:09 |
| devananda | ifarkas: hi! | 18:09 |
| ifarkas | devananda, hi, what a timing! :-) | 18:10 |
| ifarkas | devananda, I have a question regarding the vendor passthru interface. | 18:10 |
| *** iron1 has joined #openstack-ironic | 18:11 | |
| ifarkas | I am planning to add a raid management support for drac and can't decide if that should go to the management interface or to vendor passthru. I feel that for now it's a better fit for vendor passtrhu. The only problem is that it's only support post requests and asnyc processing | 18:11 |
| ifarkas | which makes a problem when I want to implement method like list_all_raid_controllers on vendor passthru | 18:11 |
| devananda | ifarkas: you're correct - that should, for now, go in vendor_passthru | 18:11 |
| devananda | ifarkas: until there is enough concensus between vendors on a standard API to move it to management interface | 18:12 |
| ifarkas | devananda, yeah, that was my thinking as well | 18:12 |
| devananda | ifarkas: so there's another thing here too -- synchronous requests to hardware are #fail | 18:12 |
| *** killer_prince is now known as lazy_prince | 18:12 | |
| devananda | issuing a GET to get any data /directly/ from the hardware is a really bad idea, as we've learned by implementing a few of those already | 18:13 |
| ifarkas | devananda, so, what would you suggest for the previous use case? | 18:13 |
| devananda | and so we've got some discussions in progress on the async API spec to do it differently | 18:13 |
| devananda | which may or may not require a /v2/ API | 18:13 |
| devananda | i haven't seen any solutions within the current API that really work | 18:14 |
| *** igordcard has joined #openstack-ironic | 18:14 | |
| devananda | ifarkas: here's an off-the-cuff possibility | 18:14 |
| *** martyntaylor1 has joined #openstack-ironic | 18:14 | |
| devananda | POST /v1/nodes/NNN/vendor_passthru?method=discover_raid_controllers | 18:14 |
| devananda | then have that populate some existing field, eg, node.extra['raid_controllers'] | 18:15 |
| *** martyntaylor has quit IRC | 18:15 | |
| ifarkas | devananda, that makes sense | 18:15 |
| devananda | the challenge will be, the user needs to wait for that update | 18:16 |
| devananda | so perhaps when the POST is received, change the field to "IN PROGRESS" or something | 18:16 |
| devananda | but this is a really hackish API :( | 18:16 |
| ifarkas | right | 18:16 |
| devananda | i feel bad even suggesting it | 18:16 |
| ifarkas | devananda, I have a trickier use case for that solution. drac supports to give back a list of supported raid levels for a given controller and a given set of disks. | 18:18 |
| ifarkas | should it go to the same place in this case too? the input and the output to node.extra? | 18:18 |
| devananda | ifarkas: so how is ironic goign to use this information? | 18:18 |
| ifarkas | devananda, it's just a helper method for the user before he issues something like the create_virtual_disk command | 18:19 |
| devananda | ifarkas: create_virtual_disk ? | 18:19 |
| ifarkas | devananda, I meant raid virtual disk | 18:20 |
| devananda | right | 18:20 |
| devananda | so, you're mapping the HW RAID commands to an API | 18:20 |
| devananda | I'd rather see a proposal for that API go through spec review | 18:20 |
| ifarkas | devananda, yeah, sort of | 18:20 |
| devananda | and expect it to take a few months, maybe more | 18:20 |
| *** martyntaylor1 has quit IRC | 18:21 | |
| devananda | so that we can get other vendors to agree that it covers their needs too | 18:21 |
| *** martyntaylor has joined #openstack-ironic | 18:21 | |
| ifarkas | yep, I was thinking about creating a spec, just wanted to check how should I use vendor_passthru | 18:21 |
| *** tzumainn has left #openstack-ironic | 18:21 | |
| ifarkas | because it's missing features like the above | 18:22 |
| ifarkas | I am not sure how to fix them | 18:22 |
| devananda | as an aside, anything you add to vendor_passthru is essentially not supported by the API, since it is not discoverable, not added to the CLI, and it is entirely driver-specific | 18:23 |
| devananda | it enables experimental features, or unique behavior of a specific vendor | 18:24 |
| devananda | raid mgmt is something that we want in the main API eventually | 18:24 |
| ifarkas | right | 18:26 |
| devananda | I think you can cobble something together using vendor_passthru here, but it doesn't support the sort of complex interactions that I would expect from a RAID mgmt API (like enumerating devices, raid level capabilities, polling BBU status, etc) | 18:27 |
| devananda | unless you stash all that data in some other field to enable async interactions | 18:27 |
| devananda | could be node.properties, for example | 18:28 |
| devananda | also, i'm thinking out loud -- so IMBW and/or change my mind :) | 18:28 |
| ifarkas | yeah, which is quite ugly on it's own | 18:28 |
| devananda | ifarkas: here's an idea | 18:29 |
| devananda | what if the driver stashes some description of capabilities on node.properties, and does thorough input validation within a driver.vendor_passthru.validate() method? | 18:31 |
| devananda | that call is synchronous | 18:31 |
| devananda | so you can return an error to the user, eg. requested virtual disk topology is not possible | 18:31 |
| devananda | as long as validate() doesn't actually touch the HW, it's fine | 18:31 |
| devananda | then POST can return 4XX error if validation fails, or 202 if validation succeeds and the request is able to be sent to the hardware | 18:32 |
| devananda | user still must poll to see the progress / status / result, but then the only difference between this API and existing power/provision/console APIs are what field the user is looking for status in | 18:33 |
| devananda | instead of {provision_state: X} they will be looking eg. at {extra: {raid_status: X} } | 18:34 |
| ifarkas | devananda, by not touching the hw, you mean it's not interacting with the drac card itself, right? | 18:35 |
| devananda | right | 18:35 |
| *** kpavel has quit IRC | 18:35 | |
| ifarkas | devananda, what would make the driver stash all the info initially? | 18:35 |
| devananda | interaction with OOB carsd (iLO, DRAC, IPMI, etc) shouldn't be done synchronously (ie, while the client is waiting for HTTP response) | 18:35 |
| *** kpavel has joined #openstack-ironic | 18:35 | |
| ifarkas | another api call on vendor passthru? | 18:36 |
| devananda | ifarkas: could be | 18:36 |
| devananda | ifarkas: I proposed a periodic task to sync hardware state as part of the asycn API changes, but those are still a ways away from becoming code, I think | 18:36 |
| ifarkas | devananda, right. ok. I like your idea. I think that's a fine middle ground for now | 18:37 |
| devananda | :) | 18:37 |
| ifarkas | devananda, thanks ;-) | 18:38 |
| devananda | ifarkas: welcome! glad I was able to be helpful :) | 18:38 |
| ifarkas | devananda, haha, definitely | 18:38 |
| devananda | ifarkas: also you may want to track https://review.openstack.org/#/c/94923/ -- async API proposal | 18:39 |
| devananda | there have been several (very different) versions proposed | 18:39 |
| devananda | Shrews has taken up the torch on that one lately, too | 18:40 |
| devananda | Shrews: you may want to read scrollback for an example use case | 18:40 |
| ifarkas | devananda, yeah, thanks, it's already on my radar | 18:41 |
| ifarkas | although I skipped the last few revisions | 18:41 |
| *** pelix has quit IRC | 18:43 | |
| NobodyCam | gah brb again | 18:46 |
| *** jbjohnso_ has joined #openstack-ironic | 18:46 | |
| *** jbjohnso__ has quit IRC | 18:49 | |
| *** bvivek has quit IRC | 18:55 | |
| *** lazy_prince is now known as killer_prince | 18:56 | |
| *** Penick has quit IRC | 19:02 | |
| *** rloo has quit IRC | 19:04 | |
| *** martyntaylor1 has joined #openstack-ironic | 19:09 | |
| *** rloo has joined #openstack-ironic | 19:10 | |
| *** martyntaylor has quit IRC | 19:11 | |
| *** Penick has joined #openstack-ironic | 19:12 | |
| iron1 | Deva, ramesh 87 uploaded a new rev of iLO deploy driver spec https://review.openstack.org/#/c/97744/ about a week ago. Can you review it? We aslo need another core reviewer to review the spec. | 19:15 |
| *** rloo has quit IRC | 19:16 | |
| *** rloo has joined #openstack-ironic | 19:17 | |
| lifeless | ok I've got a weird one | 19:26 |
| lifeless | ironic has picked a mac from a different node to register with neutron | 19:26 |
| *** malini1 has quit IRC | 19:27 | |
| NobodyCam | croupt index maybe/ | 19:27 |
| lifeless | how do you get the ports for a node? | 19:28 |
| *** malini1 has joined #openstack-ironic | 19:30 | |
| jroll | lifeless: curl /v1/nodes/uuid/ports | 19:30 |
| lifeless | ugh really, we don't do it in the CLI ? | 19:31 |
| lifeless | NobodyCam: https://bugs.launchpad.net/ironic/+bug/1342919 | 19:34 |
| rloo | lifeless: ironic node-port-list <uuid> | 19:34 |
| lifeless | oh | 19:34 |
| lifeless | no wonder I never found it | 19:34 |
| lifeless | I was doing help port-list | 19:34 |
| lifeless | NobodyCam: this is (needless to say) really breaking things :) | 19:35 |
| viktors | devananda: around? | 19:37 |
| NobodyCam | lifeless: ya... humm | 19:37 |
| lifeless | I have logs | 19:37 |
| lifeless | 4G of them sadly | 19:37 |
| jroll | lifeless: maybe, I prefer curl | 19:37 |
| jroll | typically | 19:37 |
| *** chuckC has joined #openstack-ironic | 19:38 | |
| NobodyCam | lifeless: do have an idea if this is comming from the client or the api/conductor | 19:40 |
| lifeless | NobodyCam: or the nova client? dunno yet | 19:40 |
| lifeless | NobodyCam: using the client against the api I get the right answer | 19:40 |
| NobodyCam | can you look at actual database and see if node is correctly marked there? | 19:41 |
| lifeless | I can, will do in a bit since that seems improbable as a fault since the REST API returns the right results | 19:41 |
| lifeless | and it depends on the DB | 19:41 |
| NobodyCam | ack | 19:42 |
| lifeless | see comment 2 | 19:42 |
| lifeless | ironic nova driver is asking for the wrong port | 19:42 |
| *** chuckC has quit IRC | 19:44 | |
| lifeless | hmm | 19:46 |
| lifeless | heat doesn't capture the request id | 19:46 |
| lifeless | that might be useful for debugging | 19:46 |
| lifeless | let me file a bug on that | 19:46 |
| *** martyntaylor1 has left #openstack-ironic | 19:50 | |
| lifeless | ok so the port is being looked up because the wrong node is being looked up | 19:51 |
| NobodyCam | oh that last comment is odd | 19:58 |
| NobodyCam | lifeless: how many nodes where attempting to boot? | 19:58 |
| NobodyCam | just a single? or many? | 19:58 |
| lifeless | 28 | 20:00 |
| lifeless | so we're hitting the nasty race I filed a bug about (but not as much, my workaround is enough to actually work around the scheduler issue | 20:00 |
| lifeless | I have a theory now | 20:02 |
| NobodyCam | I was going to ask which of the two nodes on the bug actually got the nova instance added to the db | 20:02 |
| *** aswadr has quit IRC | 20:02 | |
| NobodyCam | :) | 20:02 |
| lifeless | both of them serially | 20:02 |
| * NobodyCam waits | 20:02 | |
| lifeless | I think the neutron port is not being updated on a reschedule | 20:02 |
| lifeless | and not deleted | 20:02 |
| *** Penick has quit IRC | 20:05 | |
| *** Penick has joined #openstack-ironic | 20:07 | |
| lifeless | oh | 20:12 |
| lifeless | I think I know | 20:12 |
| lifeless | hypothesis: | 20:12 |
| lifeless | - nova is allocating the nwinfo before the claim happens | 20:12 |
| lifeless | - we reschedule due to failed clai | 20:12 |
| lifeless | - wrong mac boom | 20:12 |
| devananda | lifeless: correct | 20:13 |
| devananda | 20:12:28 < lifeless> - nova is allocating the nwinfo before the claim happens | 20:13 |
| *** rakesh_hs has quit IRC | 20:13 | |
| lifeless | devananda: not sure about that | 20:13 |
| lifeless | line 1315 in nova/compute/manager.py is where macs_for_instances is grabbed, and thats in the claim context | 20:13 |
| lifeless | as of two days ago | 20:14 |
| devananda | hmm | 20:14 |
| jroll | in our experience, on a reschedule: nova does release the network info correctly, but doesn't update the local cache | 20:14 |
| devananda | i'm checking now | 20:14 |
| NobodyCam | maybe a rebuild check around here: https://github.com/openstack/ironic/blob/master/ironic/drivers/modules/pxe.py#L510 | 20:14 |
| jroll | I'm fairly certain that's what we've seen... comstud fought it more than me | 20:14 |
| devananda | lifeless: https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L1965 | 20:15 |
| devananda | _build_resource is called before driver.spawn | 20:15 |
| *** ifarkas has quit IRC | 20:15 | |
| devananda | which is where the actual lock is taken in ironic | 20:15 |
| devananda | meaning that two nova processes can generate network info for the same node as they race to call driver.spawn() | 20:16 |
| lifeless | devananda: I don't think so | 20:16 |
| devananda | oh? | 20:16 |
| lifeless | there is a resource lock | 20:16 |
| lifeless | to make claims atomic | 20:16 |
| lifeless | this isn't a N-compute scenario, its one nova-compute - single node undercloud | 20:17 |
| lifeless | so that resource lock should be intact | 20:17 |
| devananda | _build_instance calls driver.macs_for_instance -- https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L1315 | 20:17 |
| lifeless | right | 20:17 |
| devananda | before calling self._allocate_network | 20:17 |
| lifeless | with rt.instance_claim(context, instance, limits): | 20:17 |
| lifeless | thats the lock | 20:17 |
| devananda | no | 20:17 |
| devananda | that's a lock on the instance in nova | 20:17 |
| lifeless | right | 20:17 |
| devananda | NOT the lock on the node in ironic | 20:17 |
| lifeless | I'm talking about nova | 20:17 |
| devananda | so two instances can be created in nova against the same node | 20:17 |
| devananda | then race | 20:18 |
| lifeless | no | 20:18 |
| devananda | why not? | 20:18 |
| lifeless | because the resource claim will fail | 20:18 |
| devananda | why? | 20:18 |
| lifeless | because the resource claim allocates all the memory on the node | 20:18 |
| lifeless | because our flavours match the nodes | 20:18 |
| lifeless | not to mention the host manager etc | 20:18 |
| devananda | resource claim is delayed update, isn't it? | 20:18 |
| lifeless | no | 20:18 |
| devananda | hm | 20:18 |
| lifeless | its atomic in the nova compute | 20:18 |
| devananda | you have >1 compute host? | 20:18 |
| lifeless | no | 20:18 |
| devananda | ok | 20:18 |
| devananda | thought you did | 20:19 |
| lifeless | and the reschedule 08:17 < lifeless> this isn't a N-compute scenario, its one nova-compute - single node undercloud | 20:19 |
| lifeless | anyhow the reschedule due to claim failures race is demonstrating that the nova-compute knows the claim details better than the scheduler | 20:19 |
| lifeless | jroll: the nwinfo cache? | 20:22 |
| jroll | lifeless: I think so, yes | 20:23 |
| viktors | devananda: ping | 20:27 |
| devananda | viktors: pong | 20:27 |
| viktors | devananda: hi! Do you have a minute to discuss patch https://review.openstack.org/#/c/107053/ (Use opportunistic approach for migration testing) ? | 20:27 |
| devananda | viktors: sure! lemme look | 20:29 |
| viktors | devananda: sure | 20:29 |
| *** malini1 has quit IRC | 20:29 | |
| devananda | viktors: it looks like this is aimed at fixing the bugs I found in oslo.db and oslo-incubator/db testing? | 20:30 |
| devananda | viktors: is this a precursor / prerequisite to moving ironic to oslo.db? | 20:30 |
| devananda | hmm, it looks like oslo.db is a dependency of this, though | 20:30 |
| viktors | devananda: correct | 20:31 |
| *** foexle has joined #openstack-ironic | 20:32 | |
| viktors | devananda: ironic.common.db code can't create a new database for each test | 20:32 |
| viktors | so we should either update common db code, or use oslo.db as a dependency | 20:33 |
| devananda | fair enough | 20:33 |
| devananda | i'm looking at the oslo.db patch too, then | 20:33 |
| lifeless | devananda: https://bugs.launchpad.net/ironic/+bug/1342919 - last comment | 20:33 |
| devananda | since that wil lhelp to understand what you're removing in 107053 | 20:33 |
| lifeless | it rescheduled once due to the scheduler race | 20:33 |
| lifeless | then tried something for 7 minutes, gave up | 20:34 |
| lifeless | then tried for 32 minutes, gave up | 20:34 |
| lifeless | I diagnosed it at 19:34 well into the last period | 20:35 |
| viktors | devananda: thanks! | 20:35 |
| *** chuckC has joined #openstack-ironic | 20:35 | |
| viktors | devananda: fyi - I suppose to go sleep soon, so I'll be able to answer to your notes in morning | 20:36 |
| devananda | viktors: ok - initial review looks fine | 20:37 |
| *** malini has joined #openstack-ironic | 20:37 | |
| devananda | viktors: i'm going to rerun my tests in a bit -- if I can confirm the bugs are closed, i'll remove my -2 and add =2 | 20:37 |
| devananda | +2 | 20:37 |
| devananda | viktors: thanks much for the work on this! | 20:37 |
| viktors | devananda: no problem, thank you for your help and patience :) | 20:38 |
| *** rloo has quit IRC | 20:41 | |
| *** viktors has quit IRC | 20:41 | |
| *** rloo has joined #openstack-ironic | 20:42 | |
| *** rloo has quit IRC | 20:42 | |
| *** romcheg1 has joined #openstack-ironic | 20:42 | |
| *** Mikhail_D_ltp has quit IRC | 20:42 | |
| *** Penick has quit IRC | 20:44 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Add versioning to Agent decommission https://review.openstack.org/106859 | 20:45 |
| *** Penick has joined #openstack-ironic | 20:45 | |
| *** rloo has joined #openstack-ironic | 20:48 | |
| *** chuckC has quit IRC | 20:56 | |
| devananda | lifeless: if you are so inclined, the iLO virtual media deploy spec LGTM. It had a previous -1 from you, which I think has been addressed. would be great to get your eyes to confirm if that's covered | 21:00 |
| devananda | NobodyCam: ditto ^ | 21:01 |
| lifeless | review # ? | 21:01 |
| devananda | 97744 | 21:01 |
| *** zdiN0bot has joined #openstack-ironic | 21:06 | |
| *** jbjohnso_ has quit IRC | 21:06 | |
| *** jdob has quit IRC | 21:07 | |
| lifeless | devananda: reviewed, my concerns are still not addressed | 21:07 |
| *** zdiN0bot has quit IRC | 21:08 | |
| *** malini has quit IRC | 21:08 | |
| *** romcheg1 has quit IRC | 21:09 | |
| NobodyCam | devananda: looking | 21:12 |
| lifeless | devananda: what do you think about https://bugs.launchpad.net/ironic/+bug/1342919/comments/8 | 21:15 |
| adam_g | lifeless, what is are your current workarounds for the scheduler issue? just https://review.openstack.org/#/c/106716/ ? | 21:24 |
| lifeless | yeah | 21:24 |
| lifeless | it reduces the severity enough to not fail out without actually landing on a node | 21:24 |
| adam_g | ah | 21:25 |
| lifeless | at least for this race | 21:26 |
| lifeless | wem | 21:26 |
| lifeless | erm | 21:26 |
| lifeless | rack | 21:26 |
| adam_g | lifeless, curious, was the node associated with the incorrect port used in any of the previous failed claims? | 21:26 |
| NobodyCam | lifeless: you saw ./nova/scheduler/filter_scheduler.py:23:1: H306 imports not in alphabetical order (time, random) on ^^^ | 21:26 |
| lifeless | NobodyCam: you saw the commit message on it ? | 21:26 |
| NobodyCam | lol /hands to face/ | 21:26 |
| NobodyCam | sorry | 21:26 |
| *** Penick has quit IRC | 21:27 | |
| *** malini has joined #openstack-ironic | 21:31 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Add versioning to Agent decommission https://review.openstack.org/106859 | 21:31 |
| *** Penick has joined #openstack-ironic | 21:32 | |
| *** romcheg1 has joined #openstack-ironic | 21:34 | |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Add versioning to Agent decommission https://review.openstack.org/106859 | 21:34 |
| openstackgerrit | A change was merged to openstack/ironic: Import a few more fixes from the Nova driver https://review.openstack.org/107214 | 21:41 |
| openstackgerrit | Josh Gachnang proposed a change to openstack/ironic-python-agent: Add versioning to Agent decommission https://review.openstack.org/106859 | 21:46 |
| lifeless | NobodyCam: https://review.openstack.org/107511 - could you perhaps take this on? | 21:49 |
| NobodyCam | devananda: i still have some consern with the add one more swap partition. ie. what if there is not enough room root_mb+swap_mb+ephemeral_mb==physical disk size | 21:51 |
| *** igordcard has quit IRC | 21:51 | |
| lifeless | devananda: one more swap partition? | 21:51 |
| NobodyCam | lifeless: lines 298-299 of https://review.openstack.org/#/c/97744/7/specs/juno/ironic-ilo-virtualmedia-driver.rst | 21:52 |
| *** matty_dubs is now known as matty_dubs|gone | 21:52 | |
| NobodyCam | line 309 removes it but.. | 21:52 |
| openstackgerrit | Ghe Rivero proposed a change to openstack/ironic: oslo.i18n migration https://review.openstack.org/105132 | 21:53 |
| lifeless | oh I missed that | 21:53 |
| * lifeless adds *another* -1 to it. | 21:53 | |
| NobodyCam | if the deploy image is a largeish custom payload that additional swap could be many GB in size | 21:55 |
| lifeless | NobodyCam: see me comment on it :) | 21:55 |
| lifeless | NobodyCam: there is no space to add that additional swap | 21:56 |
| NobodyCam | :) | 21:56 |
| lifeless | jroll: so bug found - patch up (without tests) in nova | 22:19 |
| lifeless | jroll: it may affect you too | 22:19 |
| jroll | lifeless: yeah, saw that... thanks | 22:20 |
| jroll | comstud: ^ when you have time take a look at that? | 22:20 |
| jroll | lifeless: this is the patch, yes? https://review.openstack.org/#/c/106716/ | 22:20 |
| lifeless | https://review.openstack.org/107511 | 22:20 |
| jroll | ah | 22:21 |
| comstud | ok | 22:21 |
| lifeless | the epic scheduler race is also a big bug :) | 22:21 |
| lifeless | 107511 isn't a race though, its just a plain bug | 22:21 |
| jroll | lifeless: I see, thanks | 22:21 |
| jroll | comstud: might be related to the network info cache weirdness :) | 22:21 |
| comstud | that is not the correct fix for this | 22:22 |
| comstud | I don't think | 22:22 |
| lifeless | comstud: I discussed in openstack-nova first | 22:22 |
| comstud | There's code that checks whether or not to deallocate on reschedules | 22:22 |
| comstud | did you discuss with alaski? | 22:22 |
| lifeless | yes | 22:22 |
| comstud | ok cools | 22:22 |
| lifeless | and dansmith | 22:22 |
| lifeless | uhm | 22:22 |
| lifeless | so it might be in the wrong place | 22:22 |
| lifeless | but the intent was to do it based on macs being non-None | 22:22 |
| lifeless | and macs wasn't passed around anywhere else | 22:23 |
| comstud | i had discussed this a while back with alaski | 22:23 |
| comstud | i had an item | 22:23 |
| comstud | "check to see if this is broken with ironic" | 22:23 |
| comstud | which.. I've not yet completed. | 22:23 |
| comstud | :) | 22:23 |
| lifeless | in fact, further - the next place this could be done is in the network_api | 22:23 |
| lifeless | which is definitely the wrong place | 22:23 |
| lifeless | since neutron doesn't know whether it was constrained or not | 22:23 |
| lifeless | interesting thing | 22:24 |
| lifeless | see _cleanup_allocated_networks | 22:24 |
| lifeless | vs _deallocate_network | 22:24 |
| lifeless | I wonder if the instance.system_metadata['network_allocated'] = 'False' line there | 22:24 |
| lifeless | is an existing issue on reschedules | 22:24 |
| comstud | well | 22:25 |
| *** jgrimm has quit IRC | 22:25 | |
| comstud | why i said i wasn't sure this was right was... | 22:25 |
| comstud | there was a very simple check to see if network should be deallocated | 22:25 |
| comstud | it seemed like only that check needed updated | 22:25 |
| comstud | unless.. | 22:25 |
| lifeless | I can't see that check | 22:25 |
| comstud | you've found something else that's wrong | 22:25 |
| comstud | sec | 22:25 |
| comstud | i was looking at this a month or more ago :-/ | 22:26 |
| lifeless | line 1908 | 22:26 |
| lifeless | so thats in a wrapper function | 22:26 |
| comstud | I think you're updating the OLD CODE path | 22:26 |
| comstud | :) | 22:26 |
| lifeless | which doesn't have the right info | 22:26 |
| comstud | but not positive yet | 22:26 |
| comstud | you are | 22:26 |
| comstud | _build_instance is not called anymore | 22:26 |
| comstud | deprecated. | 22:26 |
| * lifeless curses | 22:27 | |
| comstud | you want _build_and_run_instance | 22:27 |
| comstud | etc | 22:27 |
| comstud | somewhere in there is a check to see if network should be deallocated | 22:27 |
| comstud | before reschedule | 22:27 |
| comstud | (i'm looking for it too) | 22:28 |
| lifeless | I see it | 22:28 |
| lifeless | its looking at the wrong thing | 22:28 |
| comstud | oh it may be in cnoductor | 22:28 |
| comstud | conductor | 22:28 |
| comstud | you find it? | 22:28 |
| comstud | 710 if self.driver.dhcp_options_for_instance(instance): | 22:30 |
| comstud | ah yes | 22:30 |
| comstud | that's the wrong check | 22:30 |
| comstud | i recall this now | 22:31 |
| comstud | heh | 22:31 |
| comstud | i'm stuck on an old version of nova before these changes, so this all slipped my mind. | 22:31 |
| comstud | :-/ | 22:31 |
| *** Penick has quit IRC | 22:34 | |
| *** romcheg1 has quit IRC | 22:38 | |
| devananda | having really interesting conversation between Intel, NSA, and others re: trusted boot at the OSSG meetup right now | 22:48 |
| * jroll is curious | 22:49 | |
| NobodyCam | ohh | 22:49 |
| * NobodyCam wants to be fly on ye closest wall | 22:49 | |
| lifeless | comstud: v2 is up | 22:58 |
| comstud | alrighty | 22:58 |
| comstud | lifeless: you don't need any of those changes for _build_instance() | 23:01 |
| comstud | network is always deallocated before reschedule in that deprecated case | 23:02 |
| comstud | it happens in _reschedule_or_error() | 23:02 |
| comstud | when it calls self._shutdown_instance() | 23:02 |
| comstud | left a review | 23:04 |
| *** iron1 has quit IRC | 23:05 | |
| lifeless | v3 going up | 23:05 |
| NobodyCam | lol I like v3... from 20 lines to 4 (three of which are comments) | 23:07 |
| *** Penick has joined #openstack-ironic | 23:10 | |
| *** malini has quit IRC | 23:17 | |
| *** lucas-dinner has quit IRC | 23:20 | |
| *** mrda-away is now known as mrda | 23:20 | |
| mrda | Morning Ironic! | 23:20 |
| NobodyCam | good morning mrda | 23:22 |
| NobodyCam | anyone seen this b4? http://logs.openstack.org/96/107096/1/check/gate-ironic-docs/46a1647/console.html#_2014-07-15_16_14_36_487 | 23:23 |
| * NobodyCam feels like he spends too much time looking up recheck bug numbers | 23:23 | |
| mrda | devananda: ping (as per your request :) | 23:23 |
| mrda | NobodyCam: \o | 23:24 |
| mrda | NobodyCam: Can we discuss your review comment on https://review.openstack.org/#/c/107316/ ? | 23:26 |
| NobodyCam | ummm | 23:28 |
| NobodyCam | sure | 23:28 |
| devananda | mrda: hi there! | 23:28 |
| devananda | mrda: unfortunately i have about 10 minutes now | 23:28 |
| mrda | The reason for the patch is to enable 103165 to land | 23:28 |
| mrda | i.e. to address mikal and dansmith's review comments on the ironic driver | 23:29 |
| mrda | devananda: sorry, late start today due to personal | 23:29 |
| devananda | mrda: np | 23:29 |
| devananda | mrda: so - what can I help you with? | 23:29 |
| NobodyCam | ahh yes | 23:29 |
| NobodyCam | mrda: I will look again | 23:29 |
| devananda | NobodyCam: so i've been doing what mrda just proposed for a while now | 23:30 |
| devananda | the code proposed to nova should represent current tip of our tree | 23:30 |
| mrda | devananda: so the reason I didn't implement dan's comment on host_state_cls is that it would be different to the baremetal driver i.e. now we're looking at a common base class, it doesn't make a lot of sense (to me) to differ in implementation too much | 23:31 |
| devananda | which means, if we approve of 107316, we should land it ASAP and then update the corresponding proposal to nova (103165) | 23:31 |
| devananda | mrda: right - mikal's suggestion makes more sense, I think, which is what you're saying, yes? | 23:31 |
| mrda | yes, that's the plan. Once 107316 lands, I'll propose a new 103165 to reflect that and hopefully get Nova reviewer buy in | 23:31 |
| devananda | mrda: great | 23:32 |
| mrda | correct, so devananda you're happy with me saying that we'll leave host_state_cls as is since baremetal does it that way | 23:32 |
| *** malini has joined #openstack-ironic | 23:32 | |
| devananda | mrda: well, that in and of itself isn't a good argument | 23:32 |
| devananda | mrda: fix it for nova + ironic in such a way that doesn't break baremetal would be best | 23:33 |
| mrda | ...and that there's now commonality in the form of a base class between the two, as per mikal's suggestion | 23:33 |
| devananda | this existed in the baremetal driver as a hack i nthe first place. I think dan's comment is "don't preserve the hack -- do it right, this time" | 23:33 |
| *** Penick has quit IRC | 23:33 | |
| devananda | mrda: https://review.openstack.org/#/c/107316/1/ironic/nova/scheduler/base_host_manager.py | 23:34 |
| mrda | I feel like if we change it in ironic, we should change it in baremetal, for consistency | 23:34 |
| devananda | mrda: in that, new_host_state is not a class method -- it's a module method | 23:34 |
| devananda | so your comment "implement in the subclass" doesn't make sense | 23:34 |
| mrda | correct | 23:34 |
| mrda | you mean my comment here, or in code? | 23:35 |
| NobodyCam | mrda: why adding old copy writes to https://review.openstack.org/#/c/107316/1/ironic/nova/scheduler/ironic_host_manager.py | 23:35 |
| devananda | i think /want/ that to be a class method/property | 23:35 |
| devananda | which is what dan suggested | 23:35 |
| devananda | NobodyCam: see mikal's comments | 23:35 |
| devananda | NobodyCam: we incorrectly removed them in the first place | 23:35 |
| lifeless | the ironic driver doesn't seem to cache tokens | 23:35 |
| lifeless | its suppper chatty | 23:35 |
| devananda | lifeless: correct. there's a patch (from mrda!) to implement that | 23:35 |
| lifeless | review # ? | 23:35 |
| devananda | https://review.openstack.org/#/c/102695 | 23:36 |
| jroll | lifeless: https://review.openstack.org/#/c/102695/ | 23:36 |
| mrda | lifeless: so please go review https://review.openstack.org/102695 for me :) | 23:36 |
| jroll | lifeless: the bigger issue might be the... for node in icli.node_list(): icli.node_get(node) | 23:36 |
| jroll | once a minute | 23:36 |
| mrda | devananda: can you comment on https://review.openstack.org/#/c/107316/ what you'd like done with host_state_cls? | 23:37 |
| devananda | lifeless: you can also adjust the logging level of the nova.virt.ironic driver, eg. stop its DEBUG logging | 23:37 |
| devananda | without changing the level for the rest of Nova | 23:37 |
| mrda | NobodyCam: I'm not sure I understand what your question | 23:38 |
| devananda | mrda: sure. | 23:38 |
| mrda | devananda: ta | 23:38 |
| mrda | ....and any ironic reviewers out there, I'd appreciate eyeballs on https://review.openstack.org/#/c/107316 so we can progress landing the Nova Ironic Driver :) | 23:39 |
| lifeless | mrda: reviewed (-1) | 23:41 |
| lifeless | jroll: so thats way less traffic at pea | 23:42 |
| lifeless | peak | 23:42 |
| lifeless | jroll: even though its more overall | 23:42 |
| NobodyCam | mrda: I can +2 107316 if you can add a quick note in the commit message about adding the old Copyright headers back in | 23:42 |
| jroll | lifeless: what? | 23:42 |
| mrda | NobodyCam: Sure | 23:42 |
| lifeless | devananda: the chatty logs are fine, I'm talking about churn in keystone. | 23:42 |
| mrda | Give me 5 mins | 23:42 |
| *** mdorman has quit IRC | 23:43 | |
| NobodyCam | sure ... :) | 23:43 |
| lifeless | jroll: every 60 seconds it polls all nodes, but its serialised. | 23:43 |
| jroll | lifeless: it's still a lot of traffic / load | 23:43 |
| lifeless | jroll: deploy all your nodes at once, and there are dozens of calls per node, and each one of them makes a new token | 23:43 |
| lifeless | jroll: which is higher peak load | 23:43 |
| lifeless | jroll: because its running in lots of rpc threads in the compute process | 23:43 |
| lifeless | jroll: vs just one | 23:43 |
| jroll | lifeless: we've seen, when performance is degradated for whatever reason, that loop take over 60 seconds :| | 23:44 |
| jroll | but I do see your point about peak load | 23:44 |
| lifeless | jroll: I can imagine, - but i twon't reenter will it? you don't get two instances running at once ? | 23:44 |
| jroll | I think we just assume our ghetto keystone is always falling over :) | 23:44 |
| jroll | lifeless: it breaks in strange ways | 23:45 |
| lifeless | don't worry | 23:45 |
| * jroll tries to remember | 23:45 | |
| lifeless | real keystone does too | 23:45 |
| jroll | oh good | 23:45 |
| jroll | so we won't notice when we switch to real keystone | 23:45 |
| comstud | lifeless: I have a caching patch to fix the problem | 23:46 |
| devananda | mrda: comments pushed | 23:46 |
| jroll | s/patch/hack/ | 23:46 |
| comstud | get_available_nodes() is always called immediately before looping through every single node on get_available_resource() | 23:46 |
| devananda | I gotta run, but will still have gtalk for a while | 23:46 |
| comstud | eh.. patch / hack.. same thing | 23:46 |
| comstud | :) | 23:46 |
| openstackgerrit | Michael Davies proposed a change to openstack/ironic: Review fixes from Nova scheduler reviews https://review.openstack.org/107316 | 23:46 |
| mrda | thanks devananda | 23:47 |
| devananda | g'night all! see you all sporadically for the next few days of travel and conferences | 23:47 |
| mrda | have fun and see you in PODX devananda! | 23:47 |
| jroll | later deva | 23:47 |
| mrda | s/PODX/PDX | 23:47 |
| *** malini has quit IRC | 23:48 | |
| mrda | NobodyCam: I'll have a new patch up soon for your consideration - one with devananda's suggestions included | 23:51 |
| NobodyCam | yea I just read that :) | 23:51 |
| *** malini has joined #openstack-ironic | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!