| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-09-14 | |||
| 16:40:51 | opendevreview | Artom Lifshitz proposed openstack/nova stable/victoria: functional: Add live migration tests for PCI, SR-IOV servers https://review.opendev.org/c/openstack/nova/+/796907 | |
| 16:40:53 | opendevreview | Artom Lifshitz proposed openstack/nova stable/victoria: Test SRIOV port move operations with PCI conflicts https://review.opendev.org/c/openstack/nova/+/796908 | |
| 16:40:55 | opendevreview | Artom Lifshitz proposed openstack/nova stable/victoria: Update SRIOV port pci_slot when unshelving https://review.opendev.org/c/openstack/nova/+/796909 | |
| 16:40:56 | artom | For now I've just stacked on top | |
| 17:04:34 | elodilles | artom: well, it's test-only so that's good. on the other hand it is just refactoring (which is usually refused to accept as valid backport - if it's not some trivial change), thus I think it's not really "blocking". I understand that it makes backporting easier for some cases, but... | |
| 17:05:48 | artom | elodilles, so the alternative is that every backport that goes back to train (and Red Hat has to care about stable/train for a loooon time) is harder to write and review because it'll have a note about having to adjust method params or w/e | |
| 17:06:54 | elodilles | well, there will be ONE point where a patch needs modification / resolve of conflicts, but further back it can be still clean | |
| 17:07:31 | artom | elodilles, right, I meant in the sense that *every* patch going back to train will need that modification | |
| 17:07:49 | artom | As opposed to landing the test refactor and forgetting about it | |
| 17:08:45 | elodilles | let's say if we backport it till train then a patch that needs backport to stein will need the very same modification (though I understand that stein expects way less backports) | |
| 17:09:32 | artom | elodilles, ah, I see your point. We'd just be pushing back the point at which we need manual adjustments to the backport | |
| 17:10:16 | elodilles | exactly | |
| 17:10:41 | artom | I agree, I just don't know how much of an issue it will be in practice... | |
| 17:11:31 | elodilles | it depends on the amount of similar patches I think :) | |
| 17:11:42 | artom | elodilles, so, you have more experience reviewing stable branches, what would you say the ratio of RH to non-RH backports is? | |
| 17:12:28 | elodilles | definitely RH wins over non-RH, that's true :) | |
| 17:12:28 | artom | Or to put it slightly different, RH care about stable/train (queens is still a thing, though we expect way less work on it now) | |
| 17:13:05 | artom | I'm trying to tread carefully here, because this is *not* just throwing our weight around | |
| 17:13:36 | artom | If CERN (to name a random example) or Vexxhost or whoever are still running stein a need loads of backports to it, it's a different conversation | |
| 17:16:45 | elodilles | anyway, I'm not completely against backporting this, I'm just thinking as well where is the line and maybe this is somewhere there :) | |
| 17:17:44 | artom_ | But if stable/train happens to be the stable branch that's most popular by virtue of being what RH supports, it would make sense to me to make backports to train as easy as possible? <-- repeating myself in case it didn't send before my connection dropped | |
| 17:17:55 | mnaser | artom: thanks for thinking of us :) we don't care about stein anymore (thankfully :]) | |
| 17:18:12 | mnaser | in the next few weeks wallaby will be the only thing we care about | |
| 17:18:13 | artom_ | mnaser, \o/ | |
| 17:23:51 | artom | Dammit, how the crap does a new func test cause this trace: | |
| 17:23:51 | artom | 2021-09-14 13:22:47,973 ERROR [nova.api.openstack.wsgi] Unexpected exception in API method | |
| 17:23:52 | artom | Traceback (most recent call last): | |
| 17:23:52 | artom | File "/home/artom/src/nova/.tox/functional-py36/lib/python3.6/site-packages/urllib3/connectionpool.py", line 417, in _make_request | |
| 17:23:52 | artom | httplib_response = conn.getresponse(buffering=True) | |
| 17:23:52 | artom | TypeError: getresponse() got an unexpected keyword argument 'buffering' | |
| 17:23:52 | elodilles | artom: another aspect is when we backport *everything* to train, then we make it harder (or even impossible?) to other contributors (with less weight) to backport a change from train to stein as they would require to resolve a mass of conflicts | |
| 17:26:17 | artom | elodilles, valid point, though I'd counter that part of that has already been done, by virtue of the backport chain already existing for train | |
| 17:26:24 | artom | So you know you need at least those patches | |
| 17:26:50 | artom | Like, I'm willing to bite that bullet, because I think overall it saves man hours | |
| 17:27:21 | artom | I'll happily resolve conflicts for someone's stable/stein backport if it means making stable/train backporting easier | |
| 17:27:38 | artom | The question is - how do we even advertise that? | |
| 17:27:44 | artom | To let backporters know | |
| 17:27:58 | elodilles | :) valid question :) | |
| 17:28:23 | elodilles | anyway, I'll sleep on it and will review the patch tomorrow | |
| 17:29:33 | artom | elodilles, ack, thanks for the conversation :) | |
| 17:30:02 | elodilles | artom: np :) | |
| 17:30:37 | spatel | anyone has any experience with server.com to rent servers? | |
| 17:31:10 | spatel | I am planning to rent to build openstack so looking for good feedback if anyone has :) | |
| 17:33:54 | artom | elodilles, one last data point, FWIW, in the last 6 months, there's been 3 "pages" of patches to stable/train, vs 1 for stable/stein | |
| 17:34:23 | artom | https://review.opendev.org/q/project:openstack/nova+branch:stable/train+-age:6month https://review.opendev.org/q/project:openstack/nova+branch:stable/stein+-age:6month | |
| 17:34:42 | artom | So about 3 times more backports to train than stein | |
| 17:35:07 | artom | Tbh, I expected a bigger difference, I thought stein would be way less active | |
| 17:35:56 | artom | Looks like we need Vlad Gusev, he's the main stein contributor who's not RH | |
| 18:22:14 | opendevreview | Merged openstack/nova stable/ussuri: Reduce mocking in test_reject_open_redirect for compat https://review.opendev.org/c/openstack/nova/+/803094 | |
| 18:39:50 | skazi | elodilles: thx for additional info I didn't know that | |
| 18:39:58 | skazi | elodilles: thx for additional info I didn't know that :o( | |
| 18:40:04 | skazi | sean-k-mooney: thank you too! | |
| #openstack-nova - 2021-09-15 | |||
| 00:16:39 | opendevreview | melanie witt proposed openstack/nova-specs master: Re-propose Unified Limits in Nova https://review.opendev.org/c/openstack/nova-specs/+/809020 | |
| 07:35:14 | bauzas | morning folks | |
| 07:35:23 | bauzas | gibi: elodilles: good catch on the approval thing | |
| 07:35:36 | bauzas | I was about to propose the release liaison role to anyone who'd like | |
| 07:35:55 | bauzas | having a single person responsible for approving releases is a SPOF | |
| 08:51:15 | gibi | bauzas: I agree to push out such proposal. It would be nice to see somebody picks that up. If not then I can be hooked for it | |
| 08:51:41 | bauzas | I'll ask this at the next meeting | |
| 08:51:53 | bauzas | for the moment, I'm still under the water | |
| 08:51:55 | bauzas | :( | |
| 08:51:59 | gibi | bauzas: no worries | |
| 08:52:04 | bauzas | some customer is getting 4 FTEs | |
| 08:52:16 | bauzas | ... | |
| 08:52:20 | bauzas | anyway | |
| 08:52:56 | bauzas | has anyone played with this ? | |
| 08:53:09 | gibi | I'm not | |
| 09:03:11 | bauzas | ok, I found how | |
| 09:03:15 | bauzas | that's working like a charm | |
| 09:03:40 | bauzas | it just binds a new greenlet where you can access the whole process | |
| 09:04:05 | gibi | bauzas: care to document that process somewhere after you surface from the escalations? | |
| 09:04:14 | gibi | I think it would be useful | |
| 09:04:30 | bauzas | gibi: yeah that's worth of interest, I'm also testing tracemalloc module | |
| 09:04:53 | gibi | +1 for the tracemalloc too. | |
| 09:05:28 | bauzas | context (I can tell if I don't tell the whole context) : the nova-compute service is consuming unusual memory space hence a potential memory leak to identify | |
| 09:06:05 | bauzas | so the strategy is to compare memory allocations snapshots over time | |
| 09:08:15 | gibi | does tracemalloc can differentiate between allocated but not used memory and actually used memory? | |
| 09:08:42 | kashyap | Very good question | |
| 09:09:51 | kashyap | tracemalloc.get_tracemalloc_memory() | |
| 09:10:01 | bauzas | gibi: we're litterally doing baby steps here | |
| 09:10:01 | kashyap | gibi: The above _should_ get the memory usag | |
| 09:10:17 | bauzas | gibi: the idea is that you can generate snapshots | |
| 09:10:31 | bauzas | and each snapshot (like a GMR) gives you bits of memory usage | |
| 09:10:49 | bauzas | tracemalloc even provides a compare method between snapshots | |
| 09:13:21 | bauzas | kashyap: tracemalloc.get_traced_memory() only gives you the memory usage of the tracking module itself | |
| 09:27:26 | opendevreview | Lee Yarwood proposed openstack/nova master: docs: Add nova-volume volume_attachment refresh admin workflow https://review.opendev.org/c/openstack/nova/+/809161 | |
| 09:31:41 | opendevreview | Lee Yarwood proposed openstack/nova master: run-evacuate-hook: Switch to osc evacuate command and add --debug https://review.opendev.org/c/openstack/nova/+/809162 | |
| 09:42:40 | opendevreview | Merged openstack/osc-placement stable/xena: Update .gitreview for stable/xena https://review.opendev.org/c/openstack/osc-placement/+/808456 | |
| 09:45:24 | opendevreview | Merged openstack/python-novaclient stable/xena: Update .gitreview for stable/xena https://review.opendev.org/c/openstack/python-novaclient/+/808459 | |
| 09:45:25 | opendevreview | Merged openstack/python-novaclient stable/xena: Update TOX_CONSTRAINTS_FILE for stable/xena https://review.opendev.org/c/openstack/python-novaclient/+/808460 | |
| 09:54:13 | opendevreview | Merged openstack/osc-placement stable/xena: Update TOX_CONSTRAINTS_FILE for stable/xena https://review.opendev.org/c/openstack/osc-placement/+/808457 | |
| 11:44:35 | viks__ | hi, what is the use of the code: `consumer_type = 'migration'` in https://github.com/openstack/nova/blob/01eaa627b406af9336a231caea6969aae9a62033/nova/cmd/manage.py#L2550 as the next condition after this line always overrides this? am i missing something? | |
| 11:45:58 | sean-k-mooney | it will not override it if there is not instance that matches | |
| 11:46:30 | sean-k-mooney | so we asume any allocation we find that doen have a consumer type are migration and then set them to instnace if we see a corresponding instnace | |
| 11:47:23 | viks__ | sean-k-mooney: but if we check the earlier condition, we see it always seems to be setting `consumer_type = 'instance'` | |
| 11:47:45 | viks__ | https://www.irccloud.com/pastebin/XOzn2XSW/ | |
| 11:48:07 | sean-k-mooney | most allocation will be of type instance | |
| 11:49:06 | sean-k-mooney | the not is likely wrong | |
| 11:49:24 | sean-k-mooney | infact the not i think is wrong | |
| 11:49:45 | viks__ | sean-k-mooney: ok... that's what i was thinking | |