← Back to team overview

yellow team mailing list archive

Re: Test launcher script

 

On 9 April 2012 21:29, Brad Crittenden <brad.crittenden@xxxxxxxxxxxxx> wrote:
> * The script has a pdb still in it.

Wups. That's from an earlier version of the script. Apologies.

> * When running the script, the slave never came up.  'wait_for_machine' got an error as 'instance-state' was not part of the juju status output for the slave machine, as you can see at  http://paste.ubuntu.com/922372/.  Is that part of the change to the format of 'juju status' that was introduced recently.

Ah, that's odd. have you pulled down and installed the latest version
of the Python charm helpers (lp:~gmb/charm-tools/add-charm-helpers)?

> For comparison, launching a master and slave by hand (though not changing the machine type) gives this: http://paste.ubuntu.com/922365/.  The slave is still pending but at least has an instance-id and an instance-state reported.
>
> Running your script I eventually get "timeout waiting for service to start".  That is after I made a change to charmhelpers wait_for_machine to not rely on 'instance-state' existing.
>
> So, in short, I'm a bit stumped as to why a slave will not come up using your script.  Gary suggests it may be a recent change to juju.

I suspect that it might be; I certainly haven't seen this problem in
my testing, but my Juju is from Wednesday or so last week. I won't
have chance to look at it properly until Wednesday morning, sadly, but
I suspect that tweaking charmhelpers is the way to go.

... Actually, just a thought: wait_for_machine has a separate timeout
from wait_for_unit. The script tells wait_for_unit to wait for an hour
before timing out, but I think that wait_for_machine will only wait
for ~5 minutes or so... Maybe adding a call to wait_for_machine before
wait_for_unit is called in the script will serve to make sure that
everything runs smoothly.

-- 
Graham Binns | PGP Key: EC66FA7D
http://launchpad.net/~gmb


References