Last modified: 2014-09-02 11:21:21 UTC
LocalisationUpdate needs a couple of batch jobs to be set up to get the updates running periodically: 1) /home/wikipedia/bin/l10nupdate needs to run periodically (let's say every 24h initially) on a batch server, let's say hume. A cron job should be fine there. 2) All mediawiki-installation group boxes need to update their /apache/common/wmf-deployment/cache/l10n directory from the master copy in /home/wikipedia/common/wmf-deployment/cache/l10n after that script runs to update them. I'm not sure if this would be better to do in puppet, or just a regular cron job... or puppet setting up the cron job. ;) Anyway without this bit, the localizations won't update until someone does a manual sync of the whole common or PHP deployment dir.
It looks like the script generate a lot of output. Do you want said output to be emailed to you, logged somewhere or disregarded? Setting up puppet to add a cron job to each box should be pretty straightforward :)
Regarding #2, since most apache server are home-less, I am assuming you meant /usr/local/apache/common/... :) So would croning something like "rsync -a --delete 10.0.5.8::common/wmf-deployment/cache/l10n/ /usr/local/apache/common/wmf-deployment/cache/l10n/" on every apache host after the main cron-job is ran suffice?
We can stick --quiet on the PHP callout in l10nupdate to reduce the output volume. Probably should have it log to a file if there is any output though. And yes, pulling from the rsync would be ideal. :)
Both cron jobs have been setup. The update process will be logged in /var/log/l10nupdate and be over-written at every run. I have also added a date line in the update script (/home/wikipedia/bin/l10nupdate) so that the log make sense. Right now, it is setup to run the update on Hume at 2:00am and the cron jobs on the app-server cluster are setup to run at 3:00am (GMT) Let me know if there is anything else!
