Try to execute the following command (in this order)

$> lustre_unload
$> lustre_load
$> mount /work 

After executing lustre_unload, please check that indeed all Lustre modules are unloaded. If lsmod lists only the 3 QPACE modules (gs, torus and qmem), proceed with lustre_load. Otherwise, get rid of the remaining Lustre modules by executing

$> lustre_rmmod

(and probably checking again with lsmod).

In Wuppertal, the Lustre servers are prepared vor Failover - the MDS features an active/passive failover, whereas the OSS are configured for active/active failover. When there is problem with a Lustre partition, it can simply be mounted from the corresponding failover server. The following pairs are configured:

• qmdsW0 and qmdsW1
• qossW2 and qossW3

• The MDS server qmds0 dies. Then do the following on qmdsW1:

      $> mount /mnt/mgs
      $> mount /mnt/mdt-work
      

• qossW3 has no longer access to the partition /mnt/ost31. Then do the following on qossW2:

      $> mount /mnt/ost31
      

For the Jülich installation, please contact the admins.

Wuppertal - start Lustre

To start the Lustre filesystem in Wuppertal, please execute the following commands in the specified order:

• On qmdsW1:

    $> mount /mnt/mgs
    

• On qossW2:

    $> mount /mnt/ost20
    $> mount /mnt/ost21
    $> mount /mnt/ost22
    $> mount /mnt/ost23
    

• On qossW3:

    $> mount /mnt/ost30
    $> mount /mnt/ost31
    $> mount /mnt/ost32
    $> mount /mnt/ost33
    

• On qmdsW1:

    $> mount /mnt/mdt-lustre
    

Wuppertal - stopt Lustre

Either umount the partitions listed above in the reverse order, or execute the following script on qmasterW:

$> /usr/sbin/qshutdown -l

Jülich

If there are any problems with the Lustre filesystem in Jülich (server side), please contact the local admins.