fleetctl: Add restart functionality

[ryandub]

Starts and stops a unit in one command.

refs #760

This is a lot of copy+paste from the current start and stop commands. Feedback appreciated. 😄

[bcwaldon]

You actually can't skip the blocking code here, since you must guarantee that the agent stopped the unit. If you don't assert this, the restart operation might not actually cause any work to happen.

[bcwaldon]

@ryandub Rather than changing the desired state, it might work a lot better to implement this as a parallel-ssh operation. You effectively want to call systemctl restart X for each of the units, anyways, so why not just do that?

[ryandub]

@bcwaldon Thanks for the feedback. I actually didn't think about the ssh call, but you are exactly right - that would be a better way to implement this. I assume that this can be achieved just like the status and journal commands by using runCommand from

fleet/fleetctl/ssh.go

Line 189 in 019cc60

func runCommand(cmd string, machID string) (retcode int) {

.

I'm new to Go - any tips on parallelization?

Thanks!

[jonboulle]

@ryandub you can follow the same basic idea that fleetctl uses at with waitForUnitStates at the moment except it would probably look like func waitForUnitsToRestart(units []string) chan error

[ryandub]

Thanks @jonboulle - will give that a look.

[bcwaldon]

Please reopen when you've had a chance to get back to this.