| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2023-01-30 | |||
| 13:08:55 | sean-k-mooney | we might have a role for that | |
| 13:09:05 | sean-k-mooney | we do https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/ensure-rust | |
| 13:09:47 | elodilles | oh, that sounds better then \o/ | |
| 13:10:08 | elodilles | we just have to add it for failing jobs on the failing branches then i guess | |
| 13:10:29 | sean-k-mooney | add it as or too the pre playbook ya | |
| 13:10:37 | sean-k-mooney | that or have devstack do it | |
| 13:10:49 | sean-k-mooney | but the role is proably simpler | |
| 13:11:33 | sean-k-mooney | it looks like it will have rustup install the latest sable version of the tool chain | |
| 13:11:37 | sean-k-mooney | https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/ensure-rust/defaults/main.yaml | |
| 13:11:48 | sean-k-mooney | so that proably will do what we want out of the box | |
| 13:12:44 | elodilles | cool, thx, i'll try out this \o/ | |
| 14:38:34 | opendevreview | Elod Illes proposed openstack/nova master: DNM: CI test https://review.opendev.org/c/openstack/nova/+/872178 | |
| 14:57:57 | opendevreview | Elod Illes proposed openstack/nova stable/ussuri: DNM: CI test https://review.opendev.org/c/openstack/nova/+/872184 | |
| 15:03:10 | artom | So a CPU is a central processing unit | |
| 15:03:18 | artom | It's found inside computers and other electronics | |
| 15:03:27 | opendevreview | ribaudr proposed openstack/os-traits master: Add 'COMPUTE_SHARE_LOCAL_FS' https://review.opendev.org/c/openstack/os-traits/+/872185 | |
| 15:03:28 | sahid | artom: ahaha :-) | |
| 15:03:44 | sahid | bauzas: it's regarding the online/offline cpus? | |
| 15:03:49 | bauzas | yes | |
| 15:04:55 | sahid | cool, interesting work | |
| 15:05:14 | sahid | i will try to continue my review on it | |
| 15:06:06 | opendevreview | Jorge San Emeterio proposed openstack/nova master: Dividing global privsep profile https://review.opendev.org/c/openstack/nova/+/871729 | |
| 15:06:30 | sahid | i was just not sure about the files that you have added in one of your patches to simulates hw ;) | |
| 15:13:35 | opendevreview | Jorge San Emeterio proposed openstack/nova master: WIP: Moving privsep profiles to nova/__init__.py https://review.opendev.org/c/openstack/nova/+/872010 | |
| 15:23:25 | dansmith | sean-k-mooney: I think I better put this host/node consistency check in its own patch after the last one in the series, because it seems to be death by a thousand cuts on the functional tests | |
| 15:23:52 | dansmith | I also wonder if it might be better to just mock it out completely for functional | |
| 15:24:33 | dansmith | because basically every place we do a compute restart type operation, we'll fail since we don't have the node uuid thing mocked out to what it needs to be at the time | |
| 15:32:09 | opendevreview | Jorge San Emeterio proposed openstack/nova master: WIP: Moving privsep profiles to nova/__init__.py https://review.opendev.org/c/openstack/nova/+/872010 | |
| 15:46:12 | opendevreview | Aaron S proposed openstack/nova master: Add further workaround features for qemu_monitor_announce_self https://review.opendev.org/c/openstack/nova/+/867324 | |
| 15:54:47 | elodilles | sean-k-mooney: unfortunately ensure-rust does not work: https://zuul.opendev.org/t/openstack/build/f2efc09c770a448297fe90cd73191d5a | |
| 15:55:01 | elodilles | sean-k-mooney: but meanwhile i've checked and the problem is the old pip version. | |
| 15:56:02 | elodilles | sean-k-mooney: with the old pip version i was able to reproduce the 'missing rust' error, and after upgrading pip it installed cryptography fine | |
| 16:05:10 | sean-k-mooney | ack | |
| 16:09:47 | bauzas | sean-k-mooney: maybe you haven't seen but I created a Core API for the CPU states https://review.opendev.org/c/openstack/nova/+/868236/2/nova/virt/libvirt/cpu/core.py#47 | |
| 16:10:27 | bauzas | sean-k-mooney: but I can modify both modules to rather only have one Core API that would directly call the filesystem instead of calling the backed module | |
| 16:41:07 | sean-k-mooney | ill take a look later | |
| 16:41:44 | sean-k-mooney | elodilles: ack so using new pip might correct the issue provided that does not break anything with the new resolveer logic | |
| 16:45:52 | elodilles | sean-k-mooney: yes. though i have a fear that it would break something :/ | |
| 16:47:30 | elodilles | sean-k-mooney: anyway, i think the prebuilt wheel disappeared from pypi that's why we are facing this issue now :( | |
| 16:53:02 | elodilles | (or we use different pip version now than before, but I haven't find any proof for that yet) | |
| 16:53:23 | sean-k-mooney | ya i suspect it disappeared too | |
| 18:29:25 | opendevreview | Merged openstack/nova master: Persist existing node uuids locally https://review.opendev.org/c/openstack/nova/+/863918 | |
| 18:29:35 | opendevreview | Merged openstack/nova master: Fix huge-page doc https://review.opendev.org/c/openstack/nova/+/869689 | |
| 18:41:02 | dansmith | sean-k-mooney: I thought on friday you said you were happy with my answer here: https://review.opendev.org/c/openstack/nova/+/863919/12 | |
| 18:41:22 | dansmith | but I also thought you commented on it, which I don't see... do we need more on that? | |
| 18:41:51 | sean-k-mooney | i think that was in context of https://review.opendev.org/c/openstack/nova/+/863918/12 | |
| 18:42:02 | sean-k-mooney | i have not looked at the rt one since then | |
| 18:42:20 | sean-k-mooney | as in i looked at it before we talked on irc/meet | |
| 18:42:36 | sean-k-mooney | ill try and loop back too it again shortly | |
| 18:43:14 | dansmith | okay just wanted to make sure I wasn't missing some response or action | |
| 18:44:08 | sean-k-mooney | no i think we can likely proceed with that for now | |
| 18:44:11 | sean-k-mooney | https://review.opendev.org/c/openstack/nova/+/863919/12/nova/tests/unit/compute/test_resource_tracker.py#1553 | |
| 18:44:33 | dansmith | ack | |
| 18:44:44 | sean-k-mooney | i was expecting to be blocked becasue i was expecting use to block hypervior_hostname changes which we agreed should be done but is out of scope of the spec | |
| 18:45:20 | sean-k-mooney | im ok to defer that to after the current serise of patches as either a bug or something we do next cycle | |
| 18:48:47 | dansmith | I've been working all morning (and some on friday) trying to get the consistency check (i.e. "are there any nodes that claim to be mine that aren't in my list of uuids") to pass functional tests | |
| 18:49:19 | dansmith | but I'm about to slam a body part in a door so I have an excuse to stop working on it.. this stuff is such a tangled mess | |
| 18:49:20 | sean-k-mooney | i assume as a patch on the end of the seriese | |
| 18:49:47 | sean-k-mooney | i was hoping we could jsut tweak _ensure_existing_node_identity | |
| 18:49:59 | sean-k-mooney | but im guessing when we do that has test fallout | |
| 18:50:05 | sean-k-mooney | and you have been trying to resolve that | |
| 18:51:29 | dansmith | I don't want to put the logic in there specifically because that's the migration code, which we might remove one day once migrations are long out of our support envelope | |
| 18:51:42 | dansmith | but that has nothing to do with the difficulty | |
| 18:51:58 | sean-k-mooney | oh _ensure_existing_node_identity is only called there ok | |
| 18:52:00 | dansmith | I can push up what I have if you want and you can try to get things to pass | |
| 18:52:26 | sean-k-mooney | sure i can take a look at it | |
| 18:53:35 | sean-k-mooney | ill proceed with the RT change and the last two patch reviews now | |
| 18:53:54 | opendevreview | Dan Smith proposed openstack/nova master: Make resource tracker use UUIDs instead of names https://review.opendev.org/c/openstack/nova/+/863919 | |
| 18:53:55 | opendevreview | Dan Smith proposed openstack/nova master: Fixup patch for stable-compute-uuid series https://review.opendev.org/c/openstack/nova/+/871848 | |
| 18:53:55 | opendevreview | Dan Smith proposed openstack/nova master: Detect host renames and abort startup https://review.opendev.org/c/openstack/nova/+/863920 | |
| 18:53:56 | opendevreview | Dan Smith proposed openstack/nova master: TURBOWIP: Add a host/node consistency check https://review.opendev.org/c/openstack/nova/+/872204 | |
| 18:54:13 | sean-k-mooney | heh ok | |
| 18:54:22 | sean-k-mooney | i was oging to suggest not modifying the make RT one | |
| 18:54:39 | dansmith | I had to rebase to submit | |
| 18:54:58 | dansmith | it's otherwise unchanged | |
| 18:55:18 | sean-k-mooney | ya gerrit agrees so gibis +2 is still there | |
| 19:00:45 | sean-k-mooney | i see thats not exactuly what i was expecting | |
| 19:00:54 | sean-k-mooney | https://review.opendev.org/c/openstack/nova/+/872204/1/nova/compute/manager.py#1556 | |
| 19:01:38 | dansmith | well, sorry to disappoint you | |
| 19:02:32 | sean-k-mooney | i was expectign you to get teh compute node object and see if its hypervior hostname is the same as what the driver currently says it is | |
| 19:03:07 | sean-k-mooney | i think this covers the first two cases | |
| 19:03:20 | sean-k-mooney | on line 1563 | |
| 19:03:21 | dansmith | but that's not going to catch the case you were concerned about.. where we have a different (generated) uuid, so we think we should create a compute node, but we fail because of the UC on the hostname | |
| 19:03:55 | sean-k-mooney | i was concewrned about 2 thing | |
| 19:04:40 | sean-k-mooney | i think this covers test 3 and 4 https://etherpad.opendev.org/p/Stable-compute-uuid-manual-testing#L77 | |
| 19:04:58 | sean-k-mooney | im unsure about 9 and 10 | |
| 19:05:00 | sean-k-mooney | https://etherpad.opendev.org/p/Stable-compute-uuid-manual-testing#L216 | |
| 19:05:01 | dansmith | the third bullet will result in us looking up both our node and the conflicting host's node, but the uuids won't match what we expect | |
| 19:06:23 | sean-k-mooney | a change in the hostname does not maen Conf.host changes or our uuid does but i can check your code now and confirm if it resoulve all 4 test cases | |
| 19:07:27 | dansmith | I dunno what test 9 is supposed to be because there's no explanation other than "$old_hostname" so I'm not sure if this addresses that or not | |
| 19:07:37 | dansmith | I think it *does* address what I listed as things it should address, | |
| 19:07:56 | sean-k-mooney | that you change /etc/hostname form X to Y but Conf.host is hardcoded to X | |
| 19:07:57 | dansmith | the later of which is setting CONF.host to be the same for two separate compute services | |
| 19:08:22 | dansmith | that's the hypervisor_hostname thing right? this doesn't address that | |
| 19:08:36 | sean-k-mooney | thats the hypervior host name thing yes | |
| 19:08:41 | dansmith | I haven't even gotten there yet | |
| 19:08:44 | sean-k-mooney | so that what happend in old OOO | |
| 19:08:52 | sean-k-mooney | ack | |
| 19:09:34 | sean-k-mooney | tripleo started hardcodeing CONF.host before they hardcoded /etc/hostname to the same value | |
| 19:10:00 | sean-k-mooney | so 9 is basically testing that old case | |