Matthew,  just a quick thought that's not 100% related to your issue, but
is the first thing that came to mind when you said you're having issues
with a slave upon startup.  Are these dhcp machines by chance, and is it
possible that this slave is somehow getting an ip address when it wakes up
of a machine that's already connected (or is it retaining an old ip address
perhaps for a bit?)  We used to have this sort of problem with VM machines
on occasion when we'd suspend them between builds, and if I recall, putting
an ipconfig /renew fixed the issue, as it forced the local box to renew its
dhcp configuration each time it started up.

I assume you've verified on the Jenkins server that the slave shows as
disconnected there when it's down?

Scott


On Mon, Apr 28, 2014 at 1:39 PM, Matthew Forrestall <[email protected]>wrote:

> So I'm looking for advice for the following situation: We have several
> automated front end smoke tests that all run within a reverted VM on our VM
> server. Initially I thought this would work well with the Jenkins slave
> installed as a service but quickly found out that Microsoft has locked down
> the ability for services to interact with the desktop to the point now
> where it seems like running automated UI tests directly through the Jenkins
> slave as a service is not an option. Also the slave wouldn't always
> reconnect to the main Jenkins server consistently after a revert, whether I
> had it already enabled in the VM snapshot or whether I had it disabled in
> the snapshot and enabled it using psexec after the revert.
>
> At any rate, I moved to a process whereby all of the smoke test VMs have a
> snapshot that are powered off.  A batch file is installed in the Window
> startup folders that starts the Jenkins slave from the command line. The
> Jenkins workflow does the following:
>
>    1. Reverts the VM to the clean snapshot
>    2. Powers the VM on.
>    3. The batch file is executed after a short timeout and the Jenkins
>    service is started from the command line.
>
>
> This has been the most consistent approach thus far, however, one of my VM
> slaves still has a random issue connecting back to the main server. The
> error states that the slave is already connected when it isn't, since once
> the VM is reverted, that slave should get disconnected. (I will post a copy
> of the error once it fails again). We attempted to mitigate this problem by
> adding a Jenkins CLI call to explicitly disconnect the slave, and that
> hasn't improved matters.
>
> Any thoughts or advice on a better approach to starting the Jenkins slave
> in a reverted VM? I appreciate the ability to see the log output of the
> automated tests right in the Jenkins console, so I'd really like to stick
> with the Jenkins slave rather than write another tool or use a different
> workaround.
>
> --
> You received this message because you are subscribed to the Google Groups
> "Jenkins Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to