Drone (agent?) is failing to reach docker

[strk]

See ​https://dronie.osgeo.org/postgis/postgis/240

Error message is: pg-9.5: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

[strk]

Odd, one of the 3 matrix jobs did work in cloning (the pg10 one), in that build.

[robe]

As mentioned on irc. This might have to do with my upgrade to 1.0.

The job that is succeeding is the drone that runs directly on dronie-server. The other ones failing are run on ianna and debbie-docker.

It's hard to say if upgrade is the issue as lots of things were happening at same time.

The reason why a matrix job fails is each job task of a matrix is being passed to a different agent. For ow I've shut ffo the drones on ianna and debbie-docker and restarted the failing job to see if that fixes the issue.

[strk]

Shutting down ianna and debbie-docker seemed to have fixed the issue for now. Let's keep this open until we put those agents back online though, as builds are slower now...

[strk]

I'm taking it back, the problem is still not fixed. See ​https://dronie.osgeo.org/postgis/postgis/254/3/1

Any other agent connected to the server ? Can you check the logs to tell, Regina ?

[strk]

docker logs doesn't show any detail about which agents are connected from which IP addresses. The only visible information in the logs is the error of a runner with identifier "machine":"399d4d53c8f9" that starts execution and fails.

I don't know why there's no trace about any other machines.

Logs of the agent docker contain info about both successful and failing build, but w/out details of the failure, and no trace of the 399d4d53c8f9 identifier.

Where are the server and the agents configured ?

[robe]

Details here - ​https://git.osgeo.org/gitea/sac/osgeo7/wiki/Dronie-Server-container

I did turn off the others, so does seem to be some sort of dronie server agent issue.

I suppose I can revert back to before I upgraded to 1.0 version (last was rc5).

or I could just wipe out the database.

[robe]

Forgot to say not sure why it's even trying to reach the other agents when I shut them down.

[strk]

Regina can you make me an administrator of the drone-1.0 server ? Chances are there's some admin menu. How do you tell it is trying to access other agents ?

[strk]

The instructions on ​https://git.osgeo.org/gitea/sac/osgeo7/wiki/Dronie-Server-container seem to say that the server and the agent are started via lxc but while on osgeo7 machine I see them running in docker docker ps -- what's the deal ? Is there an overlap between lxc and docker ? Or are we running two services in parallel ?

[strk]

This is an interesting info to add to the wiki: ​https://github.com/drone/drone/issues/1496

[strk]

Thread on Drone support forum addressing this error: ​https://discourse.drone.io/t/cannot-connect-to-the-docker-daemon-at-unix-var-run-docker-sock-is-the-docker-daemon-running/4071

[strk]

Regina: you're missing DRONE_AGENTS_ENABLED=true to the server startup script

[strk]

Fix confirmed (was a missing DRONE_AGENTS_ENABLED variable)