Open sentry-io[bot] opened 8 months ago
I just set the DB_CONN_AGE to 180 in CL, and redployed it. Hopefully this will help.
What's a little strange is that I don't see a spike in our traffic or anything right now, so I'm not sure why this was/is happening, but it does seem to come and go.
Seems like this worked. Closing for now at least!
Sentry issue: COURTLISTENER-6PN
The issue above was Redis failing to do a DNS lookup. Lame, but it got my thinking about whether our Redis client can do connection pooling. Indeed it can, so I opened https://github.com/freelawproject/courtlistener/issues/3762 to address that, and hopefully re-fix this issue.
I think this might be fixed now that Redis and RDS are using pools. Closing for now.
What bugs me about this is I haven't noticed it correlated to any kind of traffic spike, or something else that would cause more DNS requests to be made.
Bummer that our fix to RDS and Redis didn't fix it.
Related: Sentry Issue: COURTLISTENER-74J
Filed by: @albertisfu
Also related: Sentry Issue: COURTLISTENER-74H
Filed by: @albertisfu
Related: Sentry Issue: COURTLISTENER-768
Filed by @albertisfu
This issue is affecting a lot of things:
It seems to be a network error doing a DNS lookup. Lame.
According to this post:
https://stackoverflow.com/questions/58179080/occasional-temporary-failure-in-name-resolution-while-connecting-to-aws-aurora
AWS only lets you do so many DNS requests:
I'm not sure if this will fix it, but I think we need to make fewer DNS requests across our whole system.
One way to do that is to make our DB connections more persistent by tweaking
CONN_MAX_AGE
. Currently, it's set to 60s in CL.Sentry Issue: COURTLISTENER-5S7