error decoding log message: net/http: request canceled (Client.Timeout exceeded while reading body)

[zkcrescent]

I build plugin from the codes, when I use docker logs CONTAINER_ID --follow , it occurred this error, I read some other plugin codes, basically the same as yours, so I cant find out what happened, btw, my docker version is 18.09.2, same as cli. regards

[alexandruionica]

I have the same issue with a custom logging driver which had as a base the example code here.
The only work around is to restart the Docker engine after enabling the driver and then the issue seems to go away.

I have also tried to kill the driver after enabling it in order to have Docker restart the driver only, but this doesn't lead to any improvement.

If I enable debug logging in Docker's daemon.json then it seems only part of the debug output makes it (`journalctl -u docker -f) and as soon as the restart happens I can see extra debug output like for example API calls (which could not be seen before the restart).
I'm not sure if this is the right place to report the problem but it's obscure enough that few would have encountered it and probably the main Docker project is less suited for such a report.

[clairradiant]

For anybody else running into this problem, an alternative solution is to pass --timeout=0 when calling docker plugin enable to disable http client timeouts from Docker's end. This looks to me to just be a "feature" of plugins (default request timeout of 30s) that doesn't apply to the default non-plugin log drivers.