skipped failed hosts

[dbarrosop]

runners should skip failed hosts unless stated otherwise

[nleiva]

What is a failed host?. One that you couldn't connect to?

I can add a timeout to:

if err := session.Run(r.Command); err != nil {
	result.SetErr(errors.Wrap(err, "failed to execute command"))
	jobResult <- result
	return
}

[dbarrosop]

A failed host is basically whatever the host determines that was a failure. Some things are obvious like a connection failed to establish or something like that but you could do your own thing, for instance:

var AssertionFailed error = errors.New("assertion failed")

type VerifyPreconditionsTask struct {
}

func (t *VerifyPreconditionsTask) Run(...) {
   // here we could do some asserts against a device or something
   if preConditionFailed {
       result.SetErr(AssertionFailed)  // SetErr will also store the error in the host
       jobResult <- result
   }
}

Now, as a user in your main code you could do:

gr.RunS(
   "let's test our devices meet preconditions",
   &VerifyPreconditionsTask{},
)

// Now we could filter out the devices
func okHosts(host *Host, gr *Gornir) bool {
   return host.Err() == nil
}
gr.RunS.Filter(okHosts).RunS(
  "do more stuff",
   ...
)

// Or maybe we want to do something on the hosts that failed the precondition
func hostsThatPassedPrecondition(host *Host, gr *Gornir) bool {
   return host.Err() == AssertionFailed
}
gr.RunS.Filter(hostsThatPassedPrecondition).RunS(
  "we need to do something on the hosts that failed the assertions",
   ...
)

Actually, I think this is a good enough interface without any attached magic so maybe there is nothing to do here. Maybe just write a working example.

Does this make sense?

[nleiva]

It does. I can envision running a pre-check to check the operating system version of a host, or a given capability, before sending the task. Thanks.

[dbarrosop]

Awesome, I am going to keep it open to remind us to add an example. Eventually I'd like to also use a "golden file pattern" for testing the examples.