Friday, 2022-08-05

opendevreviewwangxiyuan proposed openstack/project-config master: Update openeEuler pkg mapping
opendevreviewMerged openstack/project-config master: Update openeEuler pkg mapping
*** pojadhav is now known as pojadhav|ruck04:22
*** marios is now known as marios|ruck05:44
*** pojadhav|ruck is now known as pojadhav|rover05:54
*** bhagyashris|ruck is now known as bhagyashris06:27
*** bhagyashris_ is now known as bhagyashris07:39
*** jpena|off is now known as jpena07:59
opendevreviewMerged zuul/zuul-jobs master: upload-git-mirror: no_log around key writing
opendevreviewMerged zuul/zuul-jobs master: linters: standardise on newline at end of file
*** rlandy|out is now known as rlandy10:27
*** frenzyfriday|rover is now known as frenzyfriday|no_longer_rover10:32
*** frenzyfriday|no_longer_rover is now known as frenzyfriday10:33
*** dviroel|out is now known as dviroel11:14
opendevreviewMerged opendev/system-config master: static: move certs to group, update testing name to static99
*** jpena is now known as jpena|off11:39
fricklerinfra-root: should help detect the recent nodepool config error we had12:13
frickleralso please have a look at in order to help kolla12:13
fungifrickler: where did you see the nodepool config error? is there a traceback in our nodepool logs?12:22
opendevreviewMerged openstack/project-config master: Add nested-virt-ubuntu-jammy nodes
fricklerfungi: yes, is the fix. traceback
opendevreviewNeil Hanlon proposed openstack/project-config master: Add rockylinux 9 to nodepool
NeilHanlonfungi: pulled out the OSA change. apologies. my mind gets ahead of itself sometimes13:10
fungino worries, just figured i'd ask since the commit title said it was nodepool configuration and that was stuff for a job dashboard13:17
opendevreviewMerged opendev/system-config master: run-production-playbook-post: become when updating file time
*** dasm|off is now known as dasm14:23
clarkbfrickler: that nodepool fix failed testing. Maybe beacuse the test fixtures have the same issue (I didn't look yet)15:12
fungii was just in the process of switching gears to pull up those failed unit test builds15:16
*** dviroel is now known as dviroel|lunch15:26
*** pojadhav|rover is now known as pojadhav|out15:34
fungi"diskimage fake-image in provider real-provider not in top-level labels"15:42
fungiyep, looks like our fixtures may have the same bug in them ;)15:42
fungihmm, added that fixture in 2016, and the use case seems to be associating a diskimage with a different label name15:47
fungiso now i'm suspecting that the "fix" may be overzealous, as i think it would break this case (as evidenced by the test failure)15:48
*** marios|ruck is now known as marios|out15:52
fungino, sorry, wrong change. it came in with in 201715:53
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbHopefully we just need to wait for the api to be available there and not having problems reliably starting the api16:17
*** dviroel|lunch is now known as dviroel16:30
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkb progress. but now I need to hold a node to figure out what other valid urls are for screenshotting/testing17:19
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbfungi: re admin user creation I think the issue is django's createsuperuser command doesn't know how to accept details without prompts and interaction :/17:54
fungioh joy17:56
fungii wonder how horizon worked around that?17:56
clarkbheh and now I can't use the user I manually created from the command line without validating my email address (but the one i specified was bogus)17:56
clarkbthat is a really interesting choice considering I manually created the user on the command line via an admin command17:56
corvusclarkb:  (--noinput ?)17:58
clarkbcorvus: that foces you to provide the password via a web form aiui17:58
clarkbcorvus: thats what the upstream container images do if you set an admin user name17:58
corvusoh.  those so answers have the password in a variable17:58
clarkboh I see django accepts an env var17:58
clarkbya ok so mailman isn't exposing any of this in their container images. But we can bypass that and go straight to django17:59
clarkblet me try this17:59
corvusmaybe this is the better one
corvusone of them had a "--password" arg, the other just used the env var17:59
clarkbthe --help output from the command doesn't show --password as valid17:59
clarkbI think the env vars may be what I want though18:00
corvusah yeah, then the second one with only env vars looks more legit :)18:00
clarkbthen I need to figure out if I can force accept the email addr18:00
clarkbanother thing I'm noticing is that these containers don't really log any operational stuff to stdout/stderr. So I need to figure out where the logs are going (if they exist at all)18:03
fungii wonder how i did that for the poc... checking
clarkbok the django env vars thing appears to work18:04
clarkbbut I'm hitting email verification /me will dig18:04
fungii may not have needed to validate the address when i set it up. i guess that's what i added in step 5.1 there18:06
clarkbthey even document it there. I personally feel that is bonghits because you're operating as root against the django installation's management scripts18:07
clarkbbut ...18:07
clarkbaha! I can login to the django admin page though18:08
fungii have a feeling i never created a django superuser for the poc18:08
clarkbI think we want one to approximate our existing admin web user for mm218:09
fungiit's possible, since i was using distro-packaged mm3 and django, that the package maintscripts created a django superuser for me automatically18:09
fungimy notes mention "default debconf prompts" so one of those might have been for whether to use an existing django superuser or create a new one automatically18:11
clarkboh this gets weirder. When I attempt to login to postorius with my command line created account I get and error. I then refresh whcih reposts my login details and it shwos me the page asking me to validate my email address.18:13
clarkbIf I then login to the django admin page and refresh postorius I'm in18:13
clarkbalso when logged into the django web console as the admin account I can mark the email address as verified18:15
clarkbso I think our setup may just be to create the user. Then login and mark it verified at some point :/ at least that is workable18:15
clarkbfor configuring django to send email (to verify normal user accounts whcih we do want) I should be able to point it at localhost:25 and exim can forward that along?18:20
clarkb(the exim config may not do that currently, I don't know. but it is theoretically possible right?)18:20
fungii think our default exim config would already cover that18:36
corvusyep it should18:41
clarkbcool I wasn't sure if it would be confused between django and mm318:43
clarkbI guessi t doesn't really matter since once exim gets eamil to send all it has to do is send it18:44
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbthat patchset adds a bunch of different things beecause I don't want to do a bunch of extra CI roundtrips to test them all separately18:47
clarkband now lunch18:48
corvusclarkb: well, we do verp in exim with the existing list server, but that should only trigger if the envelope sender is *-bounces@ ; otherwise it should just behave normally18:51
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
*** dviroel is now known as dviroel|brb19:40
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
*** dviroel|brb is now known as dviroel20:56
*** dviroel is now known as dviroel|out21:12
*** dasm is now known as dasm|off22:01
clarkbfungi: one thing the spec for mm3 doesn't really cover and something I'm discovering is that the web domain and mail domains can be independent of each other. This means we can host all the domains at for web content but have the email addresses be specific. Or we can host all the different web domains for the different email domains out of the same mm3 install22:24
fungiyeah, i figured we'd do the latter for parity with the current mm2 deployment22:25
clarkbcurrently my change is putting everything under because it is easy (particularly the http -> https redirect). But I think we can pretty simply set the web domains when creating domains via the api then in theory the uwsgi proxying would just figur eout which content to serv ebased on the domain hit?22:25
clarkbya I think it will help uesrs navigate the web bits too if they don't see all the lists for everyone else22:26
clarkbI guess I should look at how to do domain specific http -> https redirects without using a bunch of domain specific vhosts22:27
fungiright. again i think i did set up two different apache vhosts for different ml domains in the poc22:27
clarkbya but I don't think it is necessary. Just need to use mod rewrite in order to redirect to the same domain but with https instead?22:28
clarkbSince I think we point at the same backend and it figures out what to serve based on the domain fed it?22:28
fungishould be able to do that with macros/vars yeah22:28
clarkb(that last bit I haven't confirmed)22:28
clarkbbut it wouldn't make senes to set a web domain value on the domains if it isn't going to figure that out for us automatically22:29
clarkbunfrotunately doesn't cover how to set the web domain details22:30
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkbI don't think that is complete, but it should get us closer22:49
clarkbone problem from a testing side is tha tI cannot easily test this due to DNS records not being in place?22:50
clarkbon monday I'll need to hold an updated test system and see if I can wrngle the vhosting into shape22:52
clarkbI'm a little worried the container image's SERVE_FROM_DOMAIN may be the thing setting the web host domain. But reading forum posts I half expect that filter vhost flag with our mail domain settings may be enough22:53
fungiclarkb: according to 5.7 of i set apache up with a default vhost and pointed the webroot at the mm3 file tree, so i think mailman itself managed to work out which domain content to serve based on the hostnames used in the requests22:59
clarkbfungi: yup and it seems the FILTER_VHOST flag must be set to true in the settings for it to do that23:03
clarkbI'm mostly wondering if there is a config item per domain we have to set to match up the domain name with the http requests23:04
clarkbbut maybe it will just work23:04
clarkbI found logs!23:06
clarkbnext ps will collect them23:06
clarkbfungi: that last ps shows a 404 there I think bceause we're filtering by vhosts and there is no loclahost vhost now23:17
opendevreviewClark Boylan proposed opendev/system-config master: WIP Add a mailman3 list server
clarkb ya I suspect that is a filter thing (and hte logs are there)23:45

Generated by 2.17.3 by Marius Gedminas - find it at!