Nomad: Support rerunning of the same job via force flag

[dvusboy]

As stated in #1575, rerunning the same task even though something external to the job-spec has changed, e.g. an artifact, does not happen in between GC cycles. The currently workaround is to trigger a GC via the HTTP API and then resubmit the job. Adding a --force flag to nomad run will make this more convenient and is more intuitive to the user.

[aroca]

Hey guys. Now the way (with downtime) would be to stop the job and then run it again?
I did not find any explanation regarding the trigger GC in docs. Perhaps you can enlighten me :)

Thanks

[tde3000]

Dunno about GC, but what I do is having an env var in the HCL called "CHANGE_ME" which I replace with sed with a random number (jenkins build number of the job that re-submits the nomad job) and because it's always different it re-evaluates and downloads the new docker image.

[transferkraM]

I'd pretty much like to have something like this as well, but especially for artifact{} I'd love to have nomad compare current runs artifacts checksums with remote checksums and trigger a re-run if checksums mismatch. e.g. using the same url in artifact{} source and allow to reschedule a job if the artifact had been updated.

Currently my ultimate goal is to maintain .nomad files with git and have continuous deployment with nomad without ever touching the jobspec files.