[StarCluster] Spot instances not removed from the queue when terminated

[Mircea Cimpoi]

Hi All,

We’ve started using StarCluster to run experiments on AWS and have
encountered some problems with the load balancer when using spot instances.

Specifically, we’re finding that the nodes are not being removed from the
queue (qconf -sel) after the spot instances are terminated and still appear
in qhost output. The jobs also appear to still be running after termination
which seems to be stopping the load balancer from adding new nodes as we
would expect.

We are using g2.2xlarge instances, and requesting 8 slots per submitted job.

Has anyone experienced similar issues?

Mircea
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/starcluster/attachments/20151124/58b6adfd/attachment.html