Follows-up T228465#5352272.
Description
Details
Related Objects
Event Timeline
@Krinkle I don't understand this task. Is the intent that the cron job in profile::openstack::eqiad1::wikitech::web be converted to a systemd timer as was done for the main cluster in profile::mediawiki::maintenance?
@bd808 As I understand it, a recent change to the TorBlock extension has made the cronjob/maintenance script obsolete.
The cron entry can be disabled/removed.
::profile::mediawiki::maintenance has a periodic job (aka systemd timer) for this recurring data load:
profile::mediawiki::periodic_job { 'mediawiki_tor_exit_node': command => '/usr/local/bin/mwscript extensions/TorBlock/maintenance/loadExitNodes.php --wiki=aawiki --force', interval => '*-*-* *:00/20:00', }
$ ssh mwmaint1002.eqiad.wmnet $ systemctl list-timers --no-pager NEXT LEFT LAST PASSED UNIT ACTIVATES Thu 2019-08-29 16:00:23 UTC 3s left Thu 2019-08-29 15:45:23 UTC 14min ago prometheus-node-exporter-apt.timer prometheus-node-exporter-apt.service Thu 2019-08-29 16:01:09 UTC 50s left Thu 2019-08-29 16:00:09 UTC 10s ago fetch_dbconfig.timer fetch_dbconfig.service Thu 2019-08-29 16:09:00 UTC 8min left Thu 2019-08-29 15:39:00 UTC 21min ago phpsessionclean.timer phpsessionclean.service Thu 2019-08-29 16:20:00 UTC 19min left Thu 2019-08-29 16:00:00 UTC 19s ago mediawiki_job_mediawiki_tor_exit_node.timer mediawiki_job_mediawiki_tor_exit_node.service Fri 2019-08-30 01:06:48 UTC 9h left Thu 2019-08-29 01:06:48 UTC 14h ago systemd-tmpfiles-clean.timer systemd-tmpfiles-clean.service Fri 2019-08-30 02:52:11 UTC 10h left Thu 2019-08-29 08:00:35 UTC 7h ago apt-daily.timer apt-daily.service Fri 2019-08-30 06:38:01 UTC 14h left Thu 2019-08-29 06:09:45 UTC 9h ago apt-daily-upgrade.timer apt-daily-upgrade.service 7 timers listed. Pass --all to see loaded but inactive timers, too.
Maybe the change was that the data is kept centrally so it only needs to be run once per cluster rather than once per wiki?
I don't understand the link with systemd timers.
It is no longer needed to run the loadExitNodes.php maintenance script. The cronjob for that maintenance script may be removed.
Implementation details: As of Gerrit changes 521020 and 524326, the MediaWiki TorBlock extension now refreshes its own data via core methods (WANCache async updates, etc.). The Memcached cache key for the data is (and was) "global", which is why the script only had to run for 1 wiki (--wiki aawiki).
EDIT: Aha, systemd timers are like cron jobs, and production has switched to those. Yeah, this isn't related to that.
EDIT (2): I assumed it was already removed in prod, but I see that's not the case. This is in just in puppet afaik, so I'll just patch it.
Change 533571 had a related patch set uploaded (by Krinkle; owner: Krinkle):
[operations/puppet@production] mediawiki: Disable loadExitNodes.php maintenance script
Change 533572 had a related patch set uploaded (by Krinkle; owner: Krinkle):
[operations/puppet@production] mediawiki: Remove loadExitNodes.php maintenance script
Excellent. This bit was what was confusing me as well and apparently I was doing a poor job of conveying that. I'll try to get you help with review + merge of the cleanup. :)
Change 533571 merged by Andrew Bogott:
[operations/puppet@production] mediawiki: Disable loadExitNodes.php maintenance script
it's because the periodic jobs class does not have an "ensure" parameter. but maybe it does not need one.
Change 537204 had a related patch set uploaded (by Dzahn; owner: Dzahn):
[operations/puppet@production] mediawiki::maintenance: remove TorExit node periodic job
Change 537204 merged by Dzahn:
[operations/puppet@production] mediawiki::maintenance: remove TorExit node periodic job
Change 533572 merged by Dzahn:
[operations/puppet@production] mediawiki: Remove loadExitNodes.php maintenance script