[StarCluster] stopping the cluster makes dismounts all EBS volumes

[MacMullan, Hugh]

Hi again Ramon:

To 'hack' the code and disable this 'feature' (detach volumes), you can comment out the 'self.detach_volumes()' line in the 'stop_cluster' definition in cluster.py:

#        self.detach_volumes()

This seems to work as expected, but maybe I'm missing something regarding the 'why' this is there at all. That would be the function (detach_volumes) that you would want to hack if you wanted to modify things to work as an option in the config, I suppose.

-Hugh

From: MacMullan, Hugh
Sent: Thursday, September 25, 2014 8:58 AM
To: 'Ramon Ramirez-Linan'
Cc: starcluster at mit.edu
Subject: RE: [StarCluster] stopping the cluster makes dismounts all EBS volumes

Hi Ramon:

If you want to change StarCluster to work differently, yes, you would have to hack the code. I think the recommended way to do this is to build a plugin, if possible,  and publicize your github code here.  If that's not possible, you could submit an 'issue' to the project: https://github.com/jtriley/StarCluster/issues. That's probably a good idea anyway, because this behavior is 'unexpected', at least to you and me. :)

All that said: in the meantime, is the workaround working for you? You could add that command to a script file, to make it more palatable to users.

Cheers,
-Hugh

From: Ramon Ramirez-Linan [mailto:rlinan at navteca.com]
Sent: Wednesday, September 24, 2014 5:32 PM
To: MacMullan, Hugh
Cc: starcluster at mit.edu<mailto:starcluster at mit.edu>
Subject: Re: [StarCluster] stopping the cluster makes dismounts all EBS volumes

Hugh,

If I wanted to implement the option of letting the user decide whether to dismount or not the volumes via config file, what would be the Starcluster way of doing this? Should i just hack the py file that deals with this?

Ramon


On Thu, Sep 18, 2014 at 10:52 AM, MacMullan, Hugh <hughmac at wharton.upenn.edu<mailto:hughmac at wharton.upenn.edu>> wrote:
Hi Ramon:

I had no idea that that StarCluster was doing that. It seems like the default behavior should be to leave any drives 'attached', and maybe have an option for the 'stop' command to detach if desired. The workaround is to use the AWS CLI to stop the cluster:

aws ec2 stop-instances --instance-ids $(for ID in $(starcluster lc hughmac 2>/dev/null | grep -e "^ *node[0-9][0-9][0-9]" -e "^ *master" | awk '{print $3}'); do echo -n "$ID "; done)

Then when you 'starcluster start mycluster' it retains the attachments and mounts (if in /etc/fstab).

Cheers,
-Hugh

From: starcluster-bounces at mit.edu<mailto:starcluster-bounces at mit.edu> [mailto:starcluster-bounces at mit.edu<mailto:starcluster-bounces at mit.edu>] On Behalf Of Ramon Ramirez-Linan
Sent: Thursday, September 18, 2014 10:22 AM
To: starcluster at mit.edu<mailto:starcluster at mit.edu>
Subject: [StarCluster] stopping the cluster makes dismounts all EBS volumes

Hi,

I have a cluster based on a custom AMI. The AMI has 2 EBS, the root one and a 1TB one.
So each node in the cluster has this 1TB EBS volume with custom data that is not shared.

The problem is that whenever I stopped the cluster with starcluster stop mycluster it dismounts the 1TB volumes and when i resume the cluster with starcluster start -x mycluster it doesnt mount the 1TB volume back.

Is there a way to tell Starcluster not to dismount the EBS when stopping?

Thanks

Ramon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/starcluster/attachments/20140925/4dcc9c30/attachment.htm