Re: NFS Mount from "External" Machine to Cluster

From: Devin Boyer <no email>
Date: Thu, 23 Apr 2015 16:35:06 -0400

Hi Vasisht,

Thanks for the info. I was aware of the volumes tag, unfortunately we'd
need to share an existing volume. Someone else suggested a plugin to me (
https://github.com/jtriley/StarCluster/issues/44#issuecomment-21766616) so
I'll check that out. I am aware of the Elastic File System, and I'm
definitely interested in playing with that once it comes out! Perhaps there
will be relevant space to author a plugin with that once it launches!

Thanks again,
-Devin

On Thu, Apr 23, 2015 at 10:53 AM, Vasisht Reddy T. <vasishtreddy_at_gmail.com>
wrote:

>
> Hi Devin,
>
> Does this NFS volume need to be simultaneously mounted on non-cluster
> instances at the same time that the cluster is running? If that’s not
> needed, then you can just use an EBS volume within the same AZ that can be
> auto mounted using the ‘VOLUMES’ tag in the config file.
> —————
>
> VOLUMES = mynfsvolume
>
> [volume mynfsvolume]
> volume_id = vol-xxxxxx
> mount_path = /mnt/nfs_dir
>
> —————————
>
> If it’s an already mounted volume, that will be have to exported by
> writing a separate plugin that runs on node addition. AWS is introducing
> an Elastic File System (EFS) which can mounted via NFS that might work for
> your case.
>
> Cheers,
> Vasisht
>
> On Wednesday, April 22, 2015 at 5:20 PM, Devin Boyer wrote:
>
> Hi everyone!
>
> I've just recently started playing with Star Cluster as we look to
> move/upgrade our current Grid Engine cluster to AWS. One sticking point
> I've run into is that in our current design, we'd like to have a directory
> on a system outside the cluster (but within the same VPC and potentially
> availability zone) that holds all of the data we will be processing on the
> cluster. I found some tools to add new NFS points from the master to the
> other nodes within the cluster, but does anyone have experience mounting an
> external NFS endpoint inside a cluster? Then from there that mount point
> could be "re-shared" to all the nodes?
>
> Basically something like this:
>
> +--------------------+
> +----------------------------------------+
> +------------------------------------------+
> | /my_nfs_dir | | /my_nfs_dir NFS mounted | |
> /my_nfs_dir NFS mounted |
> | on 1.2.3.4 | -----------> | at /mnt/nfs_dir on "master" |
> ---------> | at /mnt/nfs_dir on "node001" |
> | EC2 instance | +----------------------------------------+
> | (or other nodes) |
> +--------------------+
> +-------------------------------------------+
>
> (hopefully my diagram makes sense!)
>
> Thanks for any information you might be able to provide!
>
> -Devin Boyer
> Meteorological Engineer, WSI Inc.
> _______________________________________________
> StarCluster mailing list
> StarCluster_at_mit.edu
> http://mailman.mit.edu/mailman/listinfo/starcluster
>
>
>

-- 
* Devin Boyer* *|* *Meteorological Engineer*
* w:* 978-983-6664      *e:* devin.boyer_at_weather.com

Received on Thu Apr 23 2015 - 16:35:09 EDT