Plugin hangs Jenkins when cannot connect to docker server about docker-plugin HOT 6 CLOSED

What kind of timeouts have you set in the config ?

from docker-plugin.

The defaults:

Connection Timeout: 5
Read Timeout: 15

from docker-plugin.

Ok, so the following is happening :

• when the / page is rendered, it attempts to display each of the slaves down the RHS
• to do that, each slave is asked "are you accepting jobs"
• what we do then is check that the container exists, and is up.
  --> if the host fails to respond (or times out), we say 'no'.

Thus I suspect your host is unresponsive.

This behaviour is marginally better in the freshly minted 0.6.2, as it defers the 'check remote' to be as late as it can. It's possible we can factor that all out into the retention strategy, which is probably where it should go.

Give 0.6.2 a spin and see if it's any better.

from docker-plugin.

I'll run with 0.6.2 for a while. We'll have to see if docker server crashes again, but I can also do some tests with just shutting it down.

from docker-plugin.

So far so good, The docker server kernel panic'd and Jenkins remained responsive, though I think when I was seeing this before it was dying during a job.

from docker-plugin.

Got the docker server to kernel panic again, it occurred when the plugin attempted to start the container.

Interestingly though, before that there was a container running but no jobs were being sent to it and the docker-plugin was not removing it. I killed the container, but the docker plugin failed to make a new one reporting 500 internal server error. I restarted the whole docker service and then when the plugin attempted to start a new container, kernel panic. There is obviously some bug in docker, but that does not seem to be hanging Jenkins anymore, so I will close this ticket.

from docker-plugin.