We’re running 1.5.10, 1 admin and 3 https portals running in active/active HA.
We use the “set-config” CLI and every time we do, the https portals respond with an “upstream request timeout” for roughly a minute until after it finishes pulling its update.
The only leads I have right now is that net.core.somaxconn is being set to 4096 on our host whereas the docker-compose yaml file that DUO provides is setting it to 32768 (this is 8x4096). It looks like our host node maybe imposing the limits.
Would having the host node impose a lower somaxconn be the cause of upstream timeouts during the set-config update? I’m working on increasing the value in our environment, however it’s going to take some effort within our environment and I’m wondering what else I can check as well.