Hello guys,

we, COPR team, are very sorry for the latest service occasional unavailability. There were several performance issues that we needed to sort out:

1) High memory consumption in API1 on a copr detail - resolved by code optimization
2) Timeouts on search index updating - resolved by employing regular batch index updates instead of event-triggered ones
3) Memory consumption needed to render RubyGems monitor page - this was tackled by rewriting the code from SQLAlchemy to pure SQL

The RubyGems and PyPI package/build/monitor pages still take a long time for their delivery. Mirek Suchy implemented streamed server responses to prevent page timeouts which should alleviate the situation a little bit. We are open to any further ideas that could help us make these COPR resources more available.

After the mentioned updates, the COPR service availability should be much better. If any further issues should occur, we will react promptly.

clime