[lustre-discuss] removing zpool OSTs from a filesystem

[Anjana Kar]

After adding 4 new zpools/OSTs to an existing lustre filesystem, writes 
have
been hanging intermittently with similar messages showing up in the logs:

sd 2:0:5:0: [sdcr] Unhandled error code
sd 2:0:5:0: [sdcr] Result: hostbyte=DID_SOFT_ERROR driverbyte=DRIVER_OK
sd 2:0:5:0: [sdcr] CDB: Write(10): 2a 00 00 2b 06 70 00 00 f0 00
LustreError: 1252:0:(ost_handler.c:1775:ost_blocking_ast()) Error -2 
syncing data on lock cancel

The writes seem to recover after a while.
Has anyone seen this before? New disks are same as the existing ones, 
4TB Seagates.
Lustre version: 2.5.2

Since there is no "real" data on the OSTs, we are thinking of setting up 
a new "test"
filesystem with these 4 OSTs and a separate MDT. The new OSTs have been 
deactivated,
but they are still mounted according to zpool status on the OSS, and 
filesystem size
(df) reflects the increased size.

Should we try to  "remove" the new OSTs from the existing filesystem 
before creating the
test filesystem? Any pointers on how to do this would be greatly 
appreciated. If there
a better way to approach this problem please let me know.

Thanks in advance,
-Anjana Kar
  Pittsburgh Supercomputing Center
  kar at psc.edu