[Gluster-users] Glusterfs3.2.5 scaling
Amar Tumballi
amarts at redhat.com
Tue May 15 05:54:06 UTC 2012
On 05/15/2012 09:02 AM, alfred de sollize wrote:
> Has anybody worked with ccsm , wrf on Glusterfs? These spawn huge number
> of threads for IO.
> What are xlators to disable and what oher tune-ups so it stops giving TENC
> Al
>
>
Considering you have lot of threads doing I/O, can you try 'gluster
volume set <VOLNAME> client-io-threads enable' and see if it makes any
difference?
Regards,
Amar
>
>
>
> On Sun, May 13, 2012 at 10:54 PM, alfred de sollize <desollize at gmail.com
> <mailto:desollize at gmail.com>> wrote:
>
> Hi Amar,
> The version is in subject line - is Glusterfs-3.2.5
>
> TENC comes in clients for commands like - [cd /home] or [df -h /home]
> /home - Transport End point not connected
> it also appears in server logs,
> Attaching some part of client and server logs for one of the error
> bricks
> is some problem for application integrations also? We use Centos-6.1..
> Al
>
>
>
>
>
>
>
> On Fri, May 11, 2012 at 11:32 AM, Amar Tumballi <amarts at redhat.com
> <mailto:amarts at redhat.com>> wrote:
>
> On 05/10/2012 09:39 PM, alfred de sollize wrote:
>
> We are setting up a 180 node cluster for weather modeling. 2
> Storage
> servers with 32GB Ram each. QDR INfiniband interconnect.
> When we run iozone with 1GB perthread (128Kb blocksize) from
> 32 clients
> (2 iozone threads per client).
> The run succeeds however run fails for 64 clients and we
> start getting
> "Transport Endpoint not connected" errors.
>
> There are 10 bricks(5 from each server ) each of ~4.2TB
> making 42TB of
> export-volume that is fuse-mounted on the clients.
>
> There is no other error in the log files except for "TENC" .
>
>
> When you say log file, which files are you looking at. Also,
> which version of the glusterfs are you running ?
>
> -Amar
>
>
>
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
More information about the Gluster-users
mailing list