[Lustre-discuss] A question on time out

Andreas Dilger adilger at sun.com
Wed Oct 21 11:47:38 PDT 2009


On 2009-10-21, at 06:28, Stefano Elmopi wrote:
> I tell you that I am a novice to Lustre.
> I have the following Lustre configuration:
> OSS1
>   mount -t lustre /dev/sdc /LUSTRE_02
>   mount -t lustre 10.43.105.50 at tcp0:/lustre02 /LUSTRE_02
>
>   where /dev/sdc è a SAN partition of 100GB
>
> OSS2
>   mount -t lustre /dev/sdb /LUSTRE_02
>   mount -t lustre 10.43.105.50 at tcp0:/lustre02 /LUSTRE_02

You don't need to mount the client filesystem on top of the server
filesystem.  This will lead to some confusion.

>   where /dev/sdb è a SAN partition of 70GB
>
> for a total of a lustre02 filesystem of 170GB
>
> I have un client:
> CLIENT1
>   mount -t lustre 10.43.105.50 at tcp0:/lustre02 /LUSTRE_02
>
> At this point I launch a copy of a large file on lustre02 by CLIENT1,
> I see what node OSS is used for the copy and shutdown it.
> The current operation is suspended and here I wonder:
> how long it hangs ??
> I read the manual and I expected that the client, after a bit of time,
> try another OSS and ended the operation.
> Obviously I did not understand well.

You need to configure failover.  Lustre does not know without some
external HA software that the two disk devices on the SAN are shared.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.




More information about the lustre-discuss mailing list