Actually, I think I may have figured it out, though I haven't tracked it
down in the code. But the scenario is this:
* I launched a cluster named mynewcluster with the key dash-gsg-keypair.
* I manually killed the nodes of cluster in the AWS web console and
deleted the security groups.
* I then deleted the dash-gsg-keypair in EC2, removed it from my
config, and deleted the key from disk.
* I then launched the same cluster but with a new keypair.
At this point the listclusters and sshmaster commands exit with an error:
$ starcluster sshmaster mynewcluster
StarCluster - (
http://web.mit.edu/starcluster)
Software Tools for Academics and Researchers (STAR)
Please submit bug reports to starcluster_at_mit.edu
cli.py:1079 - ERROR - key dash-gsg-keypair not found in config
Now, if I add a dash-gsg-keypair to my starcluster config file,
listclusters works.
The reason I believe this is happening is because the terminated
instance is still present in my instances list, it still contains
reference to the dash-gsg-config key (as well as the security groups).
If this is really the case, then the command will start working without
error when the instance disappears.
I think this can happen to anyone and I do not believe its simply a
sideeffect of me trying to use sc as a library.
Austin
On 07/09/2010 01:21 PM, Austin Godber wrote:
> After I browsed the code I was under the impression that starcluster
> stored no local state beyond the config file. It had appeared that
> everything was queried from EC2 itself (for instance listing clusters).
>
> Is this correct? I am getting a KeyNotFound error for a key that I had
> deleted from prettymuch everywhere. NOTE: I am doing weird stuff e.g.
> trying to use starcluster as a library. So I may be encountering
> strange conditions.
>
> Austin
> _______________________________________________
> Starcluster mailing list
> Starcluster_at_mit.edu
> http://mailman.mit.edu/mailman/listinfo/starcluster
>
>
Received on Fri Jul 09 2010 - 13:42:02 EDT