This is a test comment, please ignore.

Thanks for fixing this! What's the difference in how we keep influxdb settings vs, say, recaptcha settings? I mean, if the latter were restored suc...

@stanhu gitlab.rb is not on this testbed is not managed so feel free to make manual changes.

Even after lowering to 4 workers, I still see OOM errors that killed both Redis and PostgreSQL:...

We have been missing data for the last few days. I found the right setting in https://performance.gitlab.net/datasources/edit/8 and added it.

@jarv The connection pool is the Rails connection pool (https://devcenter.heroku.com/articles/concurrency-and-database-connections#connection-pool)...

Then, in Rails, when we recognise a certain error signature (yet to be decided) AND we're not in a housekeeping operation, we automatically trigge...

@jarv in my testing it did not seem to matter. However, I'll test this and double check on the first one I rebuild just to be certain.

Sure, I'd imagine we can live with 4-6 workers?...

We should probably change the number of unicorn workers to account for all the extra processes on the Geo secondary....

Wrong node, sorry.