[StarCluster] New problem - instance has no alias. us-west-2b

[Lilley, John F.]

?Thought I'd follow up on this since the same problem hit again. After looking closer I realize now that the instance ID it's referring to in the 'instance has no alias' message is actually a development system which I recently assigned the running clusters security group to (@sc-bid) among other groups. Removing the security group form all instances not created by starcluster immediately corrected the problem so I assume starcluster uses the security group to query its nodes and since no alias was assigned to this foreign instance it bailed. ?


John



________________________________
From: starcluster-bounces at mit.edu <starcluster-bounces at mit.edu> on behalf of Lilley, John F. <johnbot at caltech.edu>
Sent: Wednesday, February 11, 2015 6:28 PM
To: starcluster at mit.edu
Subject: [StarCluster] New problem - instance has no alias. us-west-2b


Hi All,?


I starting running into trouble a few days ago with my Starcluster instances running in us-west-2b (Oregon) where after a day or two (most likely after the load balancer attempts to add nodes to accommodate jobs) all interaction with the cluster via starclusters management commands results in a "!!! ERROR - instance i-8ac59480 has no alias" error message. As far I know nothing has changed on my end.


Fresh starcluster instances were successfully relaunched using the most current release only to have them break again with the same error message. Trolling the starcluster mailing list archives resulted in a thread from 2013 that suggested that AWS made some temporary changes in the way userdata was handled in some regions that eventually resolved itself. http://star.mit.edu/cluster/mlarchives/1985.html

starcluster.mbox.out: Re: Multiple node cluster problems
StarCluster Mailing List Archive
Read more...<http://star.mit.edu/cluster/mlarchives/1985.html>



Hoping that someone has run into this problem and has some advice to share.


John





-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/starcluster/attachments/20150305/df984b09/attachment.htm