coreos / airlock

Minimal update/reboot orchestrator for Fedora CoreOS clusters
Apache License 2.0
39 stars 9 forks source link

server: fix leaking connections #28

Closed bjvreugdenhil closed 4 years ago

bjvreugdenhil commented 4 years ago

Hi, We encountered a problem running airlock for a longer period. Over time it stops responding and this error shows up in the logs. (multiple times) "http: Accept error: accept tcp [::]:2222: accept4: too many open files; retrying in 1s" In the container we noticed all open etcd connections:

netstat -antp
tcp 0 0 10.50.132.217:58864 10.50.31.46:2379 ESTABLISHED 1/airlock tcp 0 0 10.50.132.217:53584 10.50.31.46:2379 ESTABLISHED 1/airlock tcp 0 0 10.50.132.217:56712 10.50.31.46:2379 ESTABLISHED 1/airlock tcp 0 0 10.50.132.217:34960 10.50.31.46:2379 ESTABLISHED 1/airlock tcp 0 0 10.50.132.217:59586 10.50.31.46:2379 ESTABLISHED 1/airlock

This pull request fixes the issue