aklapper wrote on 2014-07-20 21:10:03 (UTC)

This is still a problem.

Rush wrote on 2014-08-18 18:21:41 (UTC)

assuming this ticket is an issue with search in this instance....not really a blocker for the prod instance if i understand

mattflaschen wrote on 2014-08-19 05:27:42 (UTC)

The point of this instance is to be a testbed.

If we don't understand what's causing the issue here, how will we prevent it from occurring in prod (where search will of course be very important)?

Will prod use ElasticSearch (T159: There should be a better way of communicating the message about exceeding size limit while uploading file in Phabricator)? If so, that should preferably be set up here, so we can test to make sure this issue is solved.

Rush wrote on 2014-08-19 14:58:31 (UTC)

this instance is so far off from the prod one it is an almost useless testbed for something like search, it wasn't setup the same as production with puppet and wasn't ever meant initially for the use it's getting. I understand what you are saying, but from this vantage it's not a good point

aklapper wrote on 2014-08-19 16:44:43 (UTC)

In T363#12, @mattflaschen wrote:

The point of this instance is to be a testbed.

This instance is a pretty broken though as it was a rough first setup, without proper puppetization and we face recurring issues with log files becoming too big, etc. Hence not convinced if it's really worth to investigate that much time.

If we don't understand what's causing the issue here, how will we prevent it from occurring in prod (where search will of course be very important)?

Yeah, valid question... but in the worst case this would be still fixable after Day 1 if it also becomes a problem on production.

So I am also very tempted to move this ticket to the #wikimedia_phabricator_maintenance_(after_day_1) project instead (and obviously auto-linking with brackets in project names is not supported, sigh).

Rush wrote on 2014-08-19 16:57:00 (UTC)

sounds good

mattflaschen wrote on 2014-08-20 23:11:31 (UTC)

Alright, fair enough. Interestingly, it's apparently working now anyway, so closing as Resolved. We can reopen if it recurs in production.

importing issue status

Retested again and problem does not happen: Entered "blocking reviewer" as "contains words" in the advanced Maniphest search on this instance and it lists four results, among them the expected task.