| gmann | sean-k-mooney: stephenfin on 2.99 microversion todo thing, I am generally ok with that as bug fixes in 2.99 as there is no behavior change for old microversions. +2 | 01:30 |
|---|---|---|
| opendevreview | Vasyl Saienko proposed openstack/nova master: Fix serial console for ironic https://review.opendev.org/c/openstack/nova/+/942575 | 05:07 |
| opendevreview | Rajesh Tailor proposed openstack/nova master: Add support for showing finish_time https://review.opendev.org/c/openstack/nova/+/928933 | 06:55 |
| *** ralonsoh_ is now known as ralonsoh | 09:03 | |
| Callum027 | sean-k-mooney: I've written a story on the Ceilometer side to track the changes required there to take advantage of the new patches for adding additional libvirt metadata. If you have any feedback (particularly regarding whether it'll satisfy your use case, and also my reasoning for excluding server metadata), I'd appreciate it. | 09:11 |
| Callum027 | https://storyboard.openstack.org/#!/story/2011371 | 09:11 |
| sean-k-mooney[m] | Callum027: i looked into the metadata after we talked to confirm if it was a security issue | 09:18 |
| sean-k-mooney[m] | by default ceilometer is only collecting server metadata if it starts with a “metering:” prefix | 09:19 |
| Callum027 | Yes, that's true | 09:19 |
| sean-k-mooney[m] | you can configure it to collect other keys and i think you have to opt into the collection via a config option to enable it at all | 09:20 |
| sean-k-mooney[m] | so i think it more a docs issue | 09:20 |
| sean-k-mooney[m] | i dont think we document that on the nova side | 09:20 |
| Callum027 | Just double checked, by default it looks for the "metering." prefix, and it'll include anything that starts with that (the "reserved_metadata_keys" is *in addition* to the prefix) | 09:22 |
| Callum027 | And it'll query Nova for the metadata no matter how it's configured | 09:23 |
| Callum027 | In my flavor patch for Ceilometer I've included a config option to disable that | 09:23 |
| sean-k-mooney[m] | ack | 09:24 |
| Callum027 | As far as security goes, I don't think it's a major concern if there isn't anything particularly sensitive that might go under the "metering." prefix | 09:24 |
| sean-k-mooney[m] | so i dont know the behaivor of this but there is no such thing as a reserved metadata prefix in nova | 09:25 |
| sean-k-mooney[m] | metadata in nova is exclisivly for the end user to pass info to there guest | 09:25 |
| sean-k-mooney[m] | other openstack services are not allowed to store metadata about the instance in it | 09:25 |
| sean-k-mooney[m] | so we still need to document or reconsole that in some form | 09:26 |
| Callum027 | Ceilometer itself doesn't put anything in there, it just reads anything with that prefix | 09:26 |
| Callum027 | Interestingly Ceilometer appears to expect that the server group is defined in as "metering.server_group" and it publishes that to Gnocchi if it finds it | 09:26 |
| Callum027 | I have no idea if Nova actually sets that when a server group is config | 09:27 |
| Callum027 | ured | 09:27 |
| sean-k-mooney[m] | it does not | 09:32 |
| sean-k-mooney[m] | looking at the ortinal commits that added this | 09:33 |
| sean-k-mooney[m] | i dont think ceilometer ever talks to nova about this | 09:33 |
| sean-k-mooney[m] | it appear the intent was to provide a way for end user to anotate there instance | 09:33 |
| sean-k-mooney[m] | such that metrics collected by celiometer would have those annotioation applied as lables | 09:34 |
| sean-k-mooney[m] | i.e to provide a way for the end user to query metrics about there workload using custom lables defiend in server metadata | 09:34 |
| sean-k-mooney[m] | Callum027: from a nova point of view this feature does not exist | 09:35 |
| sean-k-mooney[m] | apprently it was added for https://bugs.launchpad.net/ceilometer/+bug/1284073 | 09:39 |
| sean-k-mooney[m] | well actully no that not the one that intoduced the funciton it moved it | 09:41 |
| sean-k-mooney[m] | i think it was added in https://github.com/openstack/ceilometer/commit/00531f9de040f43c643dd843ada986cfbcf91e87#diff-41a7f81f39132a6a74adb274e6702ad8dce65ca89af0edb49cd77eeeb0875758 | 09:45 |
| adam3kijak | hi, I'm looking for any information regarding the following issue: "openstack server add volume server1 volume1", "openstack server delete server1" executed one by one results in nova-compute not executing _disconnect_volume() function due too empty block_device_mapping (https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L1755) which is updated after a short delay (race condition). Can you help me out? | 10:04 |
| Callum027 | sean-k-mooney[m]: Interesting that it tries to get the server group for instances from there then, surely there are better ways for it to get that information (though it appears that server groups aren't put into libvirt XML either) | 10:31 |
| sean-k-mooney[m] | they are not and should not be | 10:32 |
| Callum027 | Yeah, that's what I thought | 10:32 |
| sean-k-mooney[m] | the xml metadata is mainly for debugging | 10:32 |
| sean-k-mooney[m] | its not a general store for external integrations. | 10:32 |
| sean-k-mooney[m] | if something can change without reqiuring the xml to be modifed like server metadata it should not be in the xml | 10:33 |
| Callum027 | Yep, fair enough | 10:33 |
| sean-k-mooney[m] | i guess server group techinkally could be there | 10:34 |
| sean-k-mooney[m] | befause they cant change after the vms is created | 10:34 |
| sean-k-mooney[m] | but it feels a little excessive to store it | 10:34 |
| sean-k-mooney[m] | we could see what other thing but that kind of exceedign the orginal usecase | 10:35 |
| Callum027 | Yeah, I'm not proposing we include it here, it was more of an observation because I happened to be looking at what the metadata was actually being used for | 10:35 |
| Callum027 | The use case for adding more metadata than I'm proposing is tenuous. At the same time, what I'm proposing is pretty much what anyone would need in terms of determining how an instance is configured (or at least, for us anyway) | 10:38 |
| opendevreview | Rajesh Tailor proposed openstack/nova master: Add support for showing finish_time https://review.opendev.org/c/openstack/nova/+/928933 | 10:43 |
| *** ykarel_ is now known as ykarel | 10:44 | |
| opendevreview | Gaudenz Steinlin proposed openstack/nova master: Make SSH compression for remote copy configurable https://review.opendev.org/c/openstack/nova/+/943377 | 11:27 |
| masahito | gibi: hi, could you review this bug fix again? https://review.opendev.org/c/openstack/nova/+/939658 | 12:22 |
| tkajinam | masahito, looks like you have to resolve the merge conflict | 12:37 |
| tkajinam | (just in case you are not aware | 12:37 |
| opendevreview | Masahito Muroi proposed openstack/nova master: Use dict object for request_specs_dict in the _list_view https://review.opendev.org/c/openstack/nova/+/939658 | 12:47 |
| tkajinam | :-) | 12:48 |
| masahito | tkajinam: thanks for the info. The gerrit board view was quite changed, so I didn't notice the conflict :) v2.100 changed same place but I can't find the "conflict" label so I was wondering why it is. | 12:51 |
| tkajinam | masahito, I believe you could see it in the top-left (where "Active" is now shown") but the view might be a bit different according to browsers you are using | 12:52 |
| tkajinam | see https://review.opendev.org/c/openstack/requirements/+/941939 for example | 12:53 |
| tkajinam | and yeah the view has been changed | 12:53 |
| opendevreview | Merged openstack/nova-specs master: Amend the specification about the data in LiveMigrate object https://review.opendev.org/c/openstack/nova-specs/+/941417 | 13:42 |
| opendevreview | Ghanshyam proposed openstack/nova master: Testing doc job on Ubuntu Noble https://review.opendev.org/c/openstack/nova/+/935577 | 18:30 |
| opendevreview | Dan Smith proposed openstack/nova-specs master: Add one-time-use-devices spec https://review.opendev.org/c/openstack/nova-specs/+/943486 | 19:41 |
| opendevreview | Dan Smith proposed openstack/nova-specs master: Add one-time-use-devices spec https://review.opendev.org/c/openstack/nova-specs/+/943486 | 20:00 |
| opendevreview | Alan Bishop proposed openstack/nova master: Show swap_progress in attachment details https://review.opendev.org/c/openstack/nova/+/941476 | 20:28 |
| Callum027 | sean-k-mooney: Regarding your suggestion to add the image name in https://review.opendev.org/c/openstack/nova/+/942766 , I don't think image_name is actually stored in system_metadata currently. At least, in our tested on our Nova deployments it doesn't actually appear to be there. | 22:16 |
| sean-k-mooney | ya i think your right, dont try and do heroics to store it | 22:18 |
| sean-k-mooney | Callum027: if its not aviable easily not then the uuid is fine. | 22:19 |
| Callum027 | Yeah, we don't need it for our use case either so I'm happy with UUID only as well | 22:20 |
| sean-k-mooney | the name doesnet actuly matter in most case that was more a syetry thing in that we do have the falvor name | 22:31 |
| Callum027 | Come to think of it, the image name can change at any time so it's not necessarily a good idea for Nova's system metadata or the libvirt metadata to store that anyway | 23:02 |
| sean-k-mooney | it shoudl not change after its created although glance might allow it too | 23:03 |
| sean-k-mooney | ... not only do they allow it it is in the example https://docs.openstack.org/api-ref/image/v2/#id16 | 23:04 |
| sean-k-mooney | ew but ok uuid it is | 23:05 |
| opendevreview | melanie witt proposed openstack/nova master: Let oslo.limit handle retrieval of the endpoint ID https://review.opendev.org/c/openstack/nova/+/943507 | 23:43 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!