[GE users] occasional job failure - can't find user's home directory

cjf001 john.foley at motorola.com
Wed Oct 27 18:59:40 BST 2010


yes, that's probably my case, too, although I have only 2 NFS
servers that hold all the home directories. I think to "turn off"
the automounter, I'd just have to "hard-mount" the /users volume
on each of the NFS servers from all the compute nodes - so we're
talking 2 mounts, instead of hundreds - probably very doable.
However, the user volumes on each of the NFS servers are named
the same, and the automounter currently chooses between them,
so that little issue would have to be handled with some name
changes or something.

Thanks to all for their replies ! I'll have to decide whether
or not it's worth it to remove a 0.1% failure rate....

    John


Adam Tygart wrote:
> In all honesty, there are good reasons to use the automounter on a
> Beowulf cluster.
> For example, my file server is running OpenSolaris, each of my users'
> home directories falls on a different zfs "filesystem." (The reason
> for this is simpler snapshots and better quota management).
> Under your solution I would have to mount each of my 275 home
> directories on each of my 125 nodes on the off chance that a
> particular user is going to need access to those files on that node.
> This would create an unnecessary strain on my file server. Using
> automounter, only the users that are logged have their home
> directories mounted.
>
> --
> Adam
>
> On Wed, Oct 27, 2010 at 17:31, jlforrest<jlforrest at berkeley.edu>  wrote:
>> On 10/27/2010 10:19 AM, John Foley wrote:
>>> Jon -
>>>
>>> thanks for the reply - that's very interesting info. One
>>> question, though - when you say you "turned off" the
>>> automounter, I'm assuming that you also had to hard-mount
>>> the user's home directory areas, and any other NFS
>>> areas that the simulations require ?
>>
>> That's right. If the private network in a Beowulf
>> cluster has the kind of problems that the automounter
>> solves, then you have a sick cluster.
>>
>> I believe that for a standard Beowulf cluster an automounter
>> isn't necessary on the compute nodes. In environments
>> where compute nodes mount from "non-standard" file servers
>> or have other unusual features this might not be true.
>> But, in my Rocks cluster, I've had absolutely no problems
>> due to turning off the automounter.
>>
>> Jon
>>
>> ------------------------------------------------------
>> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=290496
>>
>> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
>>



-- 
###########################################################################
# John Foley                          # Location:  IL93-E1-21S            #
# IT & Systems Administration         # Maildrop:  IL93-E1-35O            #
# LV Simulation Cluster Support       #    Email: john.foley at motorola.com #
# Motorola, Inc. -  Mobile Devices    #    Phone: (847) 523-8719          #
# 600 North US Highway 45             #      Fax: (847) 523-5767          #
# Libertyville, IL. 60048  (USA)      #     Cell: (847) 460-8719          #
###########################################################################
               (this email sent using SeaMonkey on Windows)

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=290516

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list