tmary says something on ci.mozilla.org is leaking connections to ES and that's one of the reasons it's falling over. This bug is to examine our test and non-test code to see if Kitsune is leaking connections--either they're not getting closed or not getting cleaned up or something along those lines. tmary says he's using netstat to determine whether there are open connections--they'd be in an EST state. I think this should be a blocker for elastic search in production, so I'm sticking it in the next sprint.
I did the work in sprint 2012.1. The gist of it is this: pyes has an es module with an ES class in it. That class creates a connection, but has not methods for closing it. It's using urllib3 under the hood and the connection moves around a lot. Our current thinking is that SUMO is probably fine ES connection-wise. We're not keeping connections around (other than the thread-local ones). We're not doing anything fancy with connections, either. Given that, I'm going to mark this as resolved.
Status: NEW → RESOLVED
Last Resolved: 7 years ago
Resolution: --- → FIXED
Whiteboard: u=user c=questions s=2012.2 p= → u=user c=questions s=2012.1 p=
Oh, and making this a 1 pointer. I didn't spend a ton of time on it.
Whiteboard: u=user c=questions s=2012.1 p= → u=user c=questions s=2012.1 p=1
Whiteboard: u=user c=questions s=2012.1 p=1 → u=user c=questions s=2012.1 p=1 [qa-]
Target Milestone: --- → 2012-01-24
Closed as [qa-]
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.