[09:08:08] <_joe_> jbond, moritzm I'm looking at debian::autostart() and I'm not 100% sure what its behaviour is [09:08:59] <_joe_> as I read it right now, it seems that if I declare debian::autostart('foo', false) the service foo will not start after installation or upgrade [09:09:31] <_joe_> not sure about a reboot also [09:09:54] <_joe_> or am I missing something? [09:12:11] coincidentently I created https://gerrit.wikimedia.org/r/c/operations/puppet/+/708041 earlier the day :-) [09:13:00] <_joe_> moritzm: what about after a package upgrade? [09:14:29] it would also get prevented to start after an update, from the view of the postinst which runs invoke-rc the initial install and the update are identical [09:14:31] <_joe_> moritzm: also.... update-rc.d nginx disable fails on buster if nginx is not installed [09:14:35] the use case for debian::autostart [09:14:55] yes, but we don't use update-rc.d [09:14:58] <_joe_> ok then the patch for tlsproxy::instance definitely doesn't do what we wanted it to do [09:15:02] this is run via policy-rc.d [09:15:27] the use case for debian::autostart is short-lived prevention of a service start until something has run/configured/etc. [09:15:42] <_joe_> moritzm: sigh I didn't update the code [09:15:46] after which debian::autostart needs to be switched back to the default behaviour [09:15:53] <_joe_> I had an old version I guess [09:16:10] <_joe_> ok this is less baffling [09:16:19] yeah, the older version used invoke-rc.d, but it can't be used [09:16:34] <_joe_> but anyways, the tlsproxy::instance patch doesn't do what that manifest needs [09:16:44] <_joe_> I commented on it [09:16:45] since it parses daemon data from the LSB headers which are not present in a clear catch-22 [09:16:52] <_joe_> yeah [09:17:10] the policy-rc.d approach is an ugly kludge like all of sysvinit [09:17:28] but the real fix (systemd presets) only works in bullseys unfortunately [09:17:51] <_joe_> please note: systemd presets are not what tlsproxy::instance would want :) [09:18:11] <_joe_> there it's just a trick to avoid nginx from failing to install where there is already another http daemon active [09:18:35] <_joe_> we do want nginx to restart automatically after an upgrade [09:18:37] <_joe_> and at boot [09:18:52] <_joe_> it's just a critical problem at package installation times [09:19:23] ok [09:19:52] <_joe_> because all webservers at installation time want to listen on port 80 :P [10:35:46] let's move here [10:35:53] Meta is down [10:35:57] It's just not loading [10:36:10] db activity went down [10:36:11] "upstream connect error or disconnect/reset before headers. reset reason: overflow" [10:36:21] I'm getting this on enwiki [10:36:31] I'm here too [10:36:55] quick look at the network seems to be all fine [10:37:20] seems to be a very steep increase - https://grafana.wikimedia.org/d/000000550/mediawiki-application-servers?viewPanel=92&orgId=1&var-source=codfw%20prometheus%2Fops&var-cluster=appserver&var-node=mw2254 [10:38:07] I can load stuff now FYI [10:38:20] Meta loaded fine for me [10:39:03] I acked the page [10:42:45] if anyone has a moment, please update -tech/-operations topics [13:01:16] hello the Gerrit replica is currently running in read/write cause the daemon is not started with `--replica`. I could use a puppet change to be applied which brings the option back in systemd unit https://gerrit.wikimedia.org/r/c/operations/puppet/+/708102 . I can do the puppet dance & restart the service [13:02:40] hashar: in a meeting at the moment buyt will get to it after if none elses getsto it first [13:02:58] cool thanks ;) [15:31:55] {done} thank you [19:01:32] kormat: if I understand correctly, the new pc hardware in eqiad is up and runnign and tailing the old hosts, effectively cloning once we've rolled over 20 days, is that right?