OSG: fails to cancel all running pilots once all units are done

[mturilli]

$ radical-stack 
python            : 2.7.6
virtualenv        : /home/mturilli/ve/AIMES-EXPERIENCE
radical.utils     : v0.44.RC1-11-ge108e90@experiment-aimes
saga-python       : v0.44-137-g98a692ed@experiment-aimes
radical.pilot     : v0.40.1-243-g52f0a6ce@experiment-aimes

Happens often, last session with this problem:

rp.session.radical.mturilli.017223.0002

RP exits normally if I manually clean the queue from the remaining pilots.

[andre-merzky]

I consider that not relevant to 0.45, since we don't support the 'many pilots' use case there anyway, but its relevant in the 0.46 context which introduces it.

[andre-merzky]

This should be checked against the termination branch.

[ibethune]

@mturilli please test when you have the termination branches running to OSG.

[ibethune]

This is stuck behind #1241, I think.

[ibethune]

Now that #1241 is resolved (I assume from the comments that if you run with a valid project ID then you are able to run jobs), is this one still an issue?

[mturilli]

Sadly yes but I need a couple of more tests to see whether the failure I saw depended on RP.

[andre-merzky]

Matteo, should we move this one to the next release?

[andre-merzky]

did so...

[mturilli]

Thanks. Currently, OSG support seems to be low priority. I may have to run some more experiments to collect data about corner cases. In that case, I will update the ticket.