You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
- Tweak how the upstreams are setup to prevent temporary connection
failures from removing the servers from rotation.
- Allow connection retries to upstreams in the event of connection
failures.
- Enable so_keepalive on listening sockets (I don't necessarily think
this will help with the upstream keepalive issues, but is probably a
good idea, and could help with keepalive behavior to any front-facing
load balancers.
This cropped up after introducing an AWS NAT Gateway into our stack,
which closes inactive keepalive connections after 5 minutes:
https://docs.aws.amazon.com/AmazonVPC/latest/UserGuide/vpc-nat-gateway.html#nat-gateway-troubleshooting-timeout
See 18F/api.data.gov#446
0 commit comments