Docker input plugin should use context.WithTimeout

[StephanZaat]

I have around 15 dockers on a single server, so I'd prefer to use [ ] to get the status from all dockers.
However, when a docker on one of our test servers is in status "restarting" (i.e. for some reason it is not starting) Telegraf will hang on this docker and will refuse to send any output to the influx server.

I would expect Telegraf just skips a non-functioning docker. We have consul and other tools to do service discovery and checking.

[zstyblik]

@StephanZaat which version of Telegraf do you use?

[StephanZaat]

Telegraf - Version 0.12.1

[zstyblik]

This can be fixed by using any other context than context.Background() as context.Background() doesn't have any timeout. Beware, though. Metric-wise, if Docker Daemon doesn't want to talk with you or is busy, it won't talk with you. This will lead to timeouts and "gaps" in metrics. But it seems this is the reality.

[sparrc]

good point @zstyblik, I've changed the issue to reflect that. We should be using context.WithTimeout.

[sparrc]

here are the instances:

docker.go
100:    containers, err := d.client.ContainerList(context.Background(), opts)
130:    info, err := d.client.Info(context.Background())
213:    r, err := d.client.ContainerStats(context.Background(), container.ID, false)