| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-08-21 | |||
| 13:09:56 | alex_xu | gmann: enjoy your vacation! | |
| 13:10:19 | tssurya | cool, should I open a bug on the client to remove it from the option for the users ? or do we have plans on putting it on the server side | |
| 13:11:08 | mriedem | that initial novaclient change wasn't even correct, | |
| 13:11:11 | mriedem | it was later updated in https://github.com/openstack/python-novaclient/commit/fc8e5e3fe3a1164eb2e923ed599e63a2af1a4f3c | |
| 13:11:35 | alex_xu | tssurya: we have a filter called 'name', it is should be the '--instance-name' | |
| 13:11:38 | tssurya | either ways, I was trying to skip the minimal constructs for down cells for "all" filters and came across this one not abiding the rules and doing nothing except priting everything | |
| 13:12:29 | alex_xu | tssurya: sorry, that 'name' is mapping to 'display_name' also | |
| 13:12:32 | tssurya | alex_xu: yea, the documentation for those options need to be more clear to explain what means what if we are going to have both | |
| 13:12:56 | alex_xu | tssurya: yea | |
| 13:13:28 | tssurya | mriedem: oh, so you want to keep both ? | |
| 13:13:40 | mriedem | not necessarily, | |
| 13:13:46 | mriedem | clearly there is a bug in novaclient which needs to be reported | |
| 13:14:57 | tssurya | mriedem: right, I can open a bug now and we can see if this option is really useful to implement on the server side, else we can punt it. At least the documentation should be clearer for those options | |
| 13:15:23 | mriedem | so the --name filter in nova list is being mapped to filter on instance.name rather than display_name? | |
| 13:16:05 | mriedem | https://github.com/openstack/python-novaclient/commit/fc8e5e3fe3a1164eb2e923ed599e63a2af1a4f3c | |
| 13:16:06 | mriedem | oops | |
| 13:16:11 | mriedem | 'name': 'display_name', | |
| 13:16:11 | mriedem | 'image': 'image_ref', | |
| 13:16:11 | mriedem | filter_mapping = { | |
| 13:16:20 | mriedem | so we map name to display_name in the compute API code | |
| 13:16:26 | mriedem | and instance_name should map to 'name' | |
| 13:16:31 | mriedem | is what i think alex_xu was saying | |
| 13:16:42 | mriedem | if we want to support that in the server | |
| 13:16:51 | mriedem | but the client side --instance-name filter doesn't do anything today, right? | |
| 13:17:00 | alex_xu | mriedem: we have instance_name filter long time before https://github.com/openstack/nova/commit/1c90eb34085dbb69f37e2f63dea7496afabb06b3#diff-516904cc81cade24a9122ecf96707bf0R702 | |
| 13:17:20 | mriedem | (8:16:28 AM) mriedem: and instance_name should map to 'name' | |
| 13:17:20 | mriedem | right | |
| 13:17:32 | mriedem | so when was that removed? | |
| 13:20:52 | alex_xu | mriedem: probably folsom release, I see that filter in that relese, but disappear after grizzle | |
| 13:26:25 | alex_xu | mriedem: tssurya here https://review.openstack.org/#/c/10917/3 | |
| 13:27:55 | mriedem | ah ok, and forgot to remove the novaclient side of that | |
| 13:28:01 | tssurya | ah thanks | |
| 13:28:03 | mriedem | and apparently no one has noticed since folsom | |
| 13:28:51 | openstackgerrit | Chris Dent proposed openstack/nova master: Set policy_opt defaults in placement gabbi fixture https://review.openstack.org/594172 | |
| 13:29:05 | openstack | Launchpad bug 1295126 in python-novaclient "Admin only shown for args that can be used by non-admin" [Wishlist,Fix released] - Assigned to Verónica Musso (veronica-a-musso) | |
| 13:29:05 | mriedem | https://bugs.launchpad.net/python-novaclient/+bug/1295126/comments/3 | |
| 13:29:10 | mriedem | "and --instance-name has no effect for both" | |
| 13:30:13 | mriedem | tssurya: alex_xu: i'd probably just deprecate the --instance-name option in nova list, it's not done anything since essex | |
| 13:30:23 | mriedem | adding the support server-side at this point is likely a microversion | |
| 13:30:45 | alex_xu | mriedem: yea, and it is admin-only filter so we can deprecate it | |
| 13:30:48 | tssurya | mriedem: ack, I don't think its that essential a filter | |
| 13:31:14 | tssurya | its not an admin-only.. | |
| 13:33:08 | alex_xu | tssurya: instance_name field only can be see by the admin? | |
| 13:35:24 | mriedem | no, | |
| 13:35:29 | mriedem | OS-EXT-SRV-ATTR:instance_name is also shown for non-admins | |
| 13:35:35 | mriedem | it's in ExtendedServerAttributesController | |
| 13:36:37 | mriedem | oh wait no alex_xu is correc | |
| 13:36:39 | mriedem | *correct | |
| 13:36:47 | mriedem | os_compute_api:os-extended-server-attributes defaults to admin-only | |
| 13:37:06 | mriedem | https://docs.openstack.org/nova/latest/configuration/policy.html | |
| 13:38:04 | mriedem | stephenfin: see https://docs.openstack.org/nova/latest/configuration/policy.html and os_compute_api:os-extended-server-attributes - i thought we had restructured text formatting on policy option help? | |
| 13:38:15 | mriedem | maybe that's only in oslo.config option help? | |
| 13:38:17 | mriedem | efried: ? ^ | |
| 13:39:00 | efried | mriedem: Patch not merged. Lemme grab it... | |
| 13:39:12 | openstack | Launchpad bug 1295126 in python-novaclient "Admin only shown for args that can be used by non-admin" [Wishlist,Fix released] - Assigned to Verónica Musso (veronica-a-musso) | |
| 13:39:12 | tssurya | oh, well its confusing because the help for the options doesn't say its Admin only and the bug above ^ says they changed it: https://bugs.launchpad.net/python-novaclient/+bug/1295126/comments/6 | |
| 13:39:30 | mriedem | tssurya: yup, | |
| 13:39:37 | mriedem | despite that one person saying it was never even used | |
| 13:39:45 | mriedem | tssurya: so just report a bug and deprecate --instance-name | |
| 13:39:47 | mriedem | i'll +2 that | |
| 13:39:54 | tssurya | mriedem: cool | |
| 13:40:02 | mriedem | it predates gerrit so i'm not surprised it's a mess | |
| 13:40:41 | stephenfin | mriedem: Yeah, just oslo.config, I think | |
| 13:40:58 | stephenfin | though I had it in my head oslo.policy wasn't broken in the first place. Obviously not | |
| 13:41:14 | efried | - oslo.config patch to address complaint that using the rst role in help text shows up ugly in the sample: https://review.openstack.org/#/c/583064/ | |
| 13:41:14 | efried | - nova patch to twiddle a couple of options to prove it works | |
| 13:41:14 | efried | mriedem: | |
| 13:41:25 | efried | https://review.openstack.org/#/c/583025/ shoulda been that first link, sorry | |
| 13:41:42 | stephenfin | efried: I think that's a different issue | |
| 13:42:08 | stephenfin | efried: mriedem's asking why newlines and the likes in policy.help aren't being parsed | |
| 13:42:24 | stephenfin | ...in the HTML output. Your patch affects the ini output, right? | |
| 13:43:18 | tssurya | okay, so instance.display_name is name and instance.hostname is OS-EXT-SRV-ATTR:hostname and we don't care about OS-EXT-SRV-ATTR:instance_name. | |
| 13:43:22 | efried | stephenfin: the oslo.config patch, yes. | |
| 13:44:01 | efried | oh, reread what mriedem was actually saying. Yeah, I don't know about that, sorry. | |
| 13:44:09 | efried | I would have asked stephenfin :) | |
| 13:44:28 | mriedem | ha | |
| 13:44:58 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Filter out instances without a host when populating AZ https://review.openstack.org/594050 | |
| 13:45:17 | mriedem | ^ is likely an RC3 issue | |
| 13:47:51 | mriedem | regarding petr's email about install guide testing, | |
| 13:47:55 | stephenfin | mriedem: Agreed | |
| 13:48:20 | mriedem | i wonder how valid it is, or time would be saved, by starting up devstack but not enabling nova, so that you can do that manually after keystone/glance/cinder/neutron are already setup | |
| 13:48:58 | mriedem | i think the only major thing in the install guide in rocky was the placement db | |
| 13:49:29 | sean-k-mooney | mriedem: i think you will hit depency issues | |
| 13:50:00 | mriedem | on other openstack services? | |
| 13:50:06 | mriedem | or things like setting up libvirt? | |
| 13:50:34 | sean-k-mooney | well neutron would expect to be able to talks to placement for things like routed networks | |
| 13:50:48 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/rocky: Filter out instances without a host when populating AZ https://review.openstack.org/594178 | |
| 13:50:50 | mriedem | routed networks are optional and devstack doesn't set those up anyway | |
| 13:51:01 | mriedem | we definitely *should* have a ci job that uses routed networks | |
| 13:51:13 | mriedem | across a 2-node deploy where each host is in a separate aggregate | |
| 13:51:39 | mriedem | but that would require time and people that care to make sure it continues to work | |
| 13:52:12 | sean-k-mooney | in theroy devstack should be able to help i guess | |
| 13:52:45 | sean-k-mooney | mriedem: is placement installation considered part of the nova install guide | |
| 13:55:40 | dansmith | tssurya: looks like the down-cell stack needs rebasing again | |
| 13:56:00 | dansmith | presumably its review-able regardless? | |
| 13:56:01 | efried | mriedem: https://review.openstack.org/594179 <== alternative uuidsentinel impl | |
| 13:56:20 | tssurya | dansmith: yea, its ready for a first time review | |
| 13:56:26 | tssurya | still working on filtering part | |
| 13:56:32 | dansmith | okay | |
| 13:56:36 | tssurya | but would be nice to get opinions | |