I am attempting to run Gitlab in one of my Rancher environments. I strongly prefer to have my users use ssh with git commands for moving code in and out of Gitlab. This seems like it would be pretty straightforward. I have the Gitlab container working fine and I set up a Rancher load balancer to accept “TCP” "traffic on 10022 and forward it to gitlab on port 22.
This works… for a while, but then, very consistently, stops after a short time (I’m not sure if ti’s time or number of ssh requests that causes it to fail), and I have to restart the load balancer service to get it working again.
What am I doing wrong here?
Correct me if I’m wrong. After certain time, your ssh connections fails.
To fix it you have to restart LB, right ?
This seems to be related to load balancer itself rather then Gitlab.Rancher, etc.
Maybe your LB doesn’t support keep-alive and/or time outs the ssh connection to Gitlab.
Can you create new connections when old fails, or it doesn’t work as well ?
Apologies for not having replied here. We finally just have up on this. For what it worth, when I said “load balancer” here, I was referring to the Rancher load balancer service.
We finally just decided to put a host in a public subnet and give it a static IP for these gitlab ssh calls. It’s a shame, as it’s the only host we have in any of our Environments that needed to be made public. Everything else is nicely hidden in private subnets with the load balancer service being the the only gateway in. We just couldn’t get that working for this need, however.