Closed jonathonherbert closed 3 years ago
This seems sensible, especially as we have the slow log warning. Do we need an alarm for the slow logs crossing a certain threshold?
It's a good idea, I think – the project is active now so we've got 👀 on metrics, but once it's ticking along in the background these sorts of changes could drop unnoticed.
It'd also be nice to automate some load testing in CI so we have a rough idea of performance changes before we merge. We've got Gatling tests in the project that we could leverage as one way forward.
What does this change?
Some of our requests are timing out, but the service is still healthy. Timing out should be a last resort when something's gone wrong, so let's raise the limit.
This is not a substitute for a timely service – it looks like #108 has slowed us up somewhat, and we should investigate.
How to test
Test with a large document in CODE. You should not see requests time out in 10s.