Earlier  
Posted Nick Remark
#openstack-nova - 2020-05-04
14:37:01 bauzas /home/sbauza/git/openstack/nova/.tox/functional/lib/python3.7/site-packages/paste/deploy/util.py:55: DeprecationWarning: Using function/method 'Healthcheck.factory()' is deprecated: The healthcheck middleware must now be configured as an application, not as a filter
14:37:01 bauzas ahah, very interesting
14:37:02 zigo bauzas: I don't see any reason why it would be needed, but yeah, we could set /etc/nova/disable_healthcheck by default.
14:37:14 bauzas zigo: let's not overthink this
14:37:21 zigo Oh ! :)
14:37:26 bauzas this being said, see the deprecation warning above
14:37:41 zigo I've seen it as an app somewhere.
14:38:13 bauzas zigo: https://github.com/openstack/oslo.middleware/blob/2e56b4379ab5a8bd5c42e885253c81d2a5f021c2/oslo_middleware/healthcheck/__init__.py#L242-L248
14:38:22 bauzas zigo: just amending my revision
14:39:07 zigo Thanks a lot for all of this, this will really be helpful for operators.
14:40:03 zigo option httpchk GET /healthcheck
14:40:03 zigo backend novabackend
14:40:03 zigo Typically, in haproxy, you'd setup:
14:40:12 zigo If you don't, then you get crap in the nova-api.log ...
14:40:51 zigo As in, haproxy tries to open the connection, does no request, and close the connection, which is *very* hugly.
14:45:58 gmann bauzas: i feel we can keep those upgrade notes for V upgrade also as we will remove the old defaults in W cycle (along with U)- https://review.opendev.org/#/c/725226/1
14:47:24 bauzas zigo: okay, you know what ? I can't really help on the paste issue, but I'll upload my new rev
14:47:41 bauzas zigo: then you can ask gmann or other API specialists to look at your paste config
14:51:42 openstackgerrit Sylvain Bauza proposed openstack/nova master: Add a /healthcheck URL https://review.opendev.org/724684
14:52:17 bauzas zigo: okay, I left the deprecated usage but the functional test should work
14:52:42 bauzas zigo: this being said, I'll -1 because of the deprecation usage, please try to look at it
14:53:57 zigo Will do, let's first wait to see if it passes the gate.
14:55:22 bauzas zigo: it will :
14:55:32 bauzas nova.tests.functional.test_middleware.TestHealthCheckMiddleware.test_healthcheck 4.460
14:55:32 bauzas -------------------------------------------------------------------------------- -----------
14:55:32 bauzas Test id Runtime (s)
14:55:32 bauzas functional run-test: commands[1] | stestr slowest
14:55:33 bauzas __________________________________________________________________________________________________________________ summary ___________________________________________________________________________________________________________________
14:55:34 bauzas functional: commands succeeded
14:55:35 bauzas congratulations :)
14:58:18 dhellmann bauzas : o/
14:58:19 artom bauzas, zigo, also, since you're collaborating on that patch, reno and docs plz :)
14:58:21 bauzas dhellmann: thanks for joining :)
14:58:39 bauzas artom: I just helped zigo with his functest issue, nothing really important
14:58:50 bauzas now back on the RC2 holding bug
14:58:54 dhellmann I think the fix for the issue is to change the way the prelude links to the upgrade notes section, but I'm not sure of the best way to do that because you're in the middle of the RCs
14:59:10 bauzas dhellmann: ah, that's because we made an implicit target?
14:59:17 dhellmann after a final release, the rc sections are collapsed into that final version number, so the link targets change
14:59:18 dhellmann yes
14:59:27 dhellmann I'm experimenting with a fix locally
14:59:51 bauzas dhellmann: so https://review.opendev.org/#/c/725226/ wouldn't help with the issue ?
15:00:09 bauzas (sorry, got dragged from a bit)
15:00:21 bauzas gibi: you could be interested in this
15:01:01 bauzas dhellmann: we honestly used the implicit target for the prelude a lot of times
15:01:25 bauzas dhellmann: and i don't understand why it would be a problem due to https://review.opendev.org/#/q/Id9cd65877e53577bff22e408ca07bbeec4407f6e
15:02:03 gmann yeah that was i am wondering as it merged previous notes or section without comlain
15:02:27 dhellmann bauzas : no, that patch won't fix it. you might want that change anyway, but that's not the problem.
15:02:59 dhellmann bauzas : the problem is that with gmann's patch there are now 2 sections on the same page with the same title, so sphinx doesn't know which one you want to link to
15:03:21 dhellmann after the final release is tagged, they will be collapsed into 1 section, which is probably what you've seen on other releases
15:03:41 dhellmann and the problem is not really that there are 2 sections, it is that there are *links* to those sections in the text
15:03:49 jsuchome dansmith: Hi again, I rebased https://review.opendev.org/#/c/572805 and https://review.opendev.org/#/c/574301/ , could I get some reviews, please?
15:03:51 dhellmann so sphinx has to resolve where those links should go
15:04:13 bauzas dhellmann: looking at the paste you provided
15:05:58 bauzas okay, I *think* I understand
15:06:14 bauzas by merging gmann's patch, we're creating a new section
15:06:19 dhellmann right
15:06:45 dansmith jsuchome: I had feedback, so addressing that is necessary in addition to just rebasing it
15:07:13 dhellmann sheesh, there are a lot of changes in nova :-) generating the release notes takes ages
15:10:18 gmann dhellmann: should not merge section happen when it is release say rc2 version relese? and till then anything new goes into unreleased section instead of ussuri release section ?
15:10:31 dhellmann I think I have a fix, give me a sec to rebase
15:11:38 gmann dhellmann: ok, thanks
15:12:59 gmann bauzas: dhellmann may be i could have edited the existing reno instead of new so that it would conflict but that just hides this issue does not solve.
15:13:31 gmann bauzas: gibi i was searching and could not find any such issue (where new reno added in rc) for train and stein rc.
15:17:11 gibi honestly I can only agree what you found with the reno problem but I cannot give to that discussion. I hope dhellmann's fix will help
15:19:41 openstackgerrit Doug Hellmann proposed openstack/nova stable/ussuri: use more distinct link references in release notes https://review.opendev.org/725301
15:19:41 openstackgerrit Doug Hellmann proposed openstack/nova stable/ussuri: Add nova-status upgrade check and reno for policy new defaults https://review.opendev.org/725146
15:19:52 dhellmann let's see if ^ helps
15:22:33 jsuchome dansmith: well, yeah, I'm not sure how to address it. Are you suggesting some different path then using this formerly deprecated entry point? As I'm not the creator of the original code I would rather not rework it completely...
15:22:44 dhellmann gibi : sphinx is like a compiler & linker for docs instead of for code. If there are 2 sections on a page with the same name, that's like having 2 functions in a module with the same name. Sphinx can't figure out which one you want to link to.
15:23:07 gibi dhellmann: thanks. and also thanks for fixing it
15:23:28 dansmith jsuchome: yes, I'm not okay with undeprecating that hook mechanism just to add something in-tree for it
15:23:51 dhellmann this is only related to reno because it's assembling the notes files into rst for sphinx. the same error could have happened if the text had been written by hand
15:30:58 gibi dhellmann: so your patch can fix it as it changes the references current pointing to the ambiguous title to point to some specific title that is not duplicated by gmann's recent patch
15:31:13 dhellmann that's right
15:31:16 gmann dhellmann: thanks for fixing. and that did not complain in original change because Sphinx already merged the Upgrade Section - https://review.opendev.org/#/c/725301/1/releasenotes/notes/ussuri-prelude-4b96f1244cefcdf4.yaml
15:31:25 dhellmann the patch isn't quite right (it failed locally) but it's a different error
15:31:45 bauzas sorry was in a meeting
15:32:30 gmann dhellmann: any way Sphinx can error or warning for using implicit links at least for same link in same page? just for future so that it can be fixed in original change itself.
15:33:04 dhellmann good question. I don't know.
15:33:21 dhellmann there's a `reno lint` command, maybe that's a check it could do?
15:33:49 bauzas I'm a bit afraid of https://review.opendev.org/#/c/725301/ because it would be a bit more difficult for people to provide their own reno notes
15:34:01 gmann dhellmann: ok. let me try sometime later.
15:34:16 bauzas but we're close to RC2 so I wouldn't be picky
15:34:47 dhellmann another possible improvement for reno is to have it automatically add unique anchors for each file so that the prelude could have used those anchor names
15:35:12 dhellmann the problem with that is that the same note can appear in multiple series (backports) so you can't just use the filename
15:36:05 bauzas dhellmann: gmann: gibi: if you wanna get MHO, I'd just prefer to https://review.opendev.org/#/c/725146/2/releasenotes/notes/bug-1875418-0df3198e36530ec7.yaml to just amend an existing upgrade note instead of creating this YAML file
15:36:13 gmann dhellmann: yeah, asking for explicit anchor is much safe in case of backporting things.
15:36:34 bauzas we could leave some placeholder YAML file for backports
15:36:56 bauzas this would be simplier to manage rather than explicit sphinx targets
15:36:59 dhellmann bauzas : the whole point of reno is you don't have to do that sort of thing
15:37:03 bauzas I know :(
15:37:06 gmann bauzas: but that does not solve the whole things, same issue can occur for any new reno while backport
15:37:26 dhellmann another fix would be to remove the links in the prelude entirely
15:37:49 gibi bauzas: would you like to do that change on master first to keep the reno (even if it is ignored) in sync between branches?
15:38:10 bauzas gibi: which change ?N
15:38:41 gibi bauzas: ammending the existing upgrade reno instead of adding a new one
15:38:52 openstackgerrit Doug Hellmann proposed openstack/nova stable/ussuri: use more distinct link references in release notes https://review.opendev.org/725301
15:38:53 openstackgerrit Doug Hellmann proposed openstack/nova stable/ussuri: Add nova-status upgrade check and reno for policy new defaults https://review.opendev.org/725146
15:39:02 dhellmann ok, I think ^ solves it
15:39:44 dhellmann the 2nd problem was that the text of the links was the same, too, so `title <url1>`_ and `title <url2>`_ was causing a conflict. The fix was to use double underscore so sphinx doesn't try to "define" a link with the title

Earlier   Later