Error handling for docker swarm mode

[tanyadegurechaff]

Description

Fixes #1100

[emilevauge]

@tanyadegurechaff
You introduced a memory leak (you need to close the new channel).
Could you explain in details how your code fixes the issue? Isn't this issue Swarm related instead ?
Ping @vdemeester

[tanyadegurechaff]

This makes backoff operation return error when an error occurs, and the operation will be retried.

[tanyadegurechaff]

Could you explain in detail how a memory leak comes about?

[beldpro-ci]

AFAIK not closing a channel does not introduce a memory leak by itself (as one could worry in regards to files). As I see it's just a mechanism for communicating that nothing more will be sent there, right? (I'd really like to know if I got something wrong).

[emilevauge]

@tanyadegurechaff @beldpro-ci After reading more in details, indeed, it does not introduce any memory leak in this case. But it is still better to defer close the channel, especially because it's inside a go routine. We had many leaks in the past due to opened channels. Other than that, seems great :)

[tanyadegurechaff]

Thank you for your feedbacks. @beldpro-ci @emilevauge

[ldez]

why you put the defer here?
I think you must move this line under the line 156.

[tanyadegurechaff]

The receiver at line 185 blocks the execution until the channel is closed or received an error.

[emilevauge]

I agree with @tanyadegurechaff on this. But it would be better to use this syntax instead:

if err, ok := <-errChan; ok {
  return err
} else {
  // channel closed
  ...
}

WDYT?

[tanyadegurechaff]

I agree, but I got an error from golint:

provider/docker/docker.go:187:13: if block ends with a return statement, so drop this else and outdent its block (move short variable declaration to its own line if necessary)

[ldez]

just do:

if err, ok := <-errChan; ok {
  return err
}
// channel closed
// ...

[ldez]

Due to SemaphoreCI, I close and reopen the PR.

[emilevauge]

Thanks @tanyadegurechaff
LGTM !