You need to cope with incorrect keypair. Cause you can have some instances started with different keypairs.<br><br><div class="gmail_quote">2009/9/18 Justin Riley <span dir="ltr"><<a href="mailto:jtriley@mit.edu">jtriley@mit.edu</a>></span><br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">-----BEGIN PGP SIGNED MESSAGE-----<br>
Hash: SHA1<br>
<br>
Hi Alex,<br>
<br>
Thank you for using StarCluster! I'd like to invite you to join the<br>
mailing list at some point. I've cc'd the list so that others can view<br>
the response, I hope you don't mind.<br>
<br>
Regarding your bug report, this is something I'm aware of and am working<br>
to fix in the next release of StarCluster.<br>
<br>
Currently, StarCluster has almost no accounting for what nodes were<br>
started by StarCluster previously. For now, the software assumes that<br>
the only EC2 instances you have in AWS are those started by StarCluster<br>
AND that only one StarCluster has been started. Only under these<br>
circumstances will the -m and -n options will work correctly.<br>
<br>
This is clearly NOT optimal. I plan to implement a mechanism for proper<br>
cluster accounting (including multiple cluster support):<br>
<br>
My idea thus far is to:<br>
<br>
1. Iterate through all EC2 instances (regardless of AMI) looking for a<br>
file on the instance that indicates that it's a master node.<br>
<br>
2. This file will contain all of the slave instances associated with<br>
that master and thus help to eliminate the need for exhaustive search<br>
(ie it will let me reduce the number of instances to search as I<br>
encounter each master node).<br>
<br>
3. This information will then be cached (for performance) and used for<br>
the -l, -m, and -n options.<br>
<br>
4. I will likely add a -u option will update the local cache.<br>
<br>
5. Anytime you run the -s or -t options, the cache will also be updated.<br>
<br>
6. Along with this I will be adding a -c option for specifying which<br>
cluster you would like these options (ie -m, -n, -l, -t etc) to operate on.<br>
<br>
This is the direction I'm heading as of now, but I would certainly<br>
entertain any other ideas on how to properly account for clusters<br>
launched by StarCluster.<br>
<br>
I'll keep you posted as I make progress on this.<br>
<br>
Thanks again for using StarCluster!<br>
<br>
~Justin<br>
<br>
<br>
> Hello,<br>
><br>
> I've several running instances in AWS, and start 2 cluster instances. If<br>
> I try to invoke "starcluster -m" it is trying to make ssh connection to<br>
> previously running instances - not by starcluster<br>
><br>
> ---<br>
> Best regards,<br>
> Alexander Gorban<br>
-----BEGIN PGP SIGNATURE-----<br>
Version: GnuPG v2.0.11 (GNU/Linux)<br>
Comment: Using GnuPG with Mozilla - <a href="http://enigmail.mozdev.org" target="_blank">http://enigmail.mozdev.org</a><br>
<br>
iEYEARECAAYFAkqzyRYACgkQ4llAkMfDcrnV2QCfbvzxVhgY62NtBBiYiIX4hpFI<br>
z4EAn3dhPaKhAzOR5bXoUTBmXSa7V78A<br>
=T7zN<br>
-----END PGP SIGNATURE-----<br>
</blockquote></div><br>---<br>Best regards,<br>Alexander Gorban<br>