Replication of a backup of a 1.2PB filesystem

PeterChilds

Newcomer
Joined
Jan 29, 2020
Messages
2
Reaction score
0
Points
0
We are attempting to replicate a Spectrum Protect backup of our Spectrum Scale file system, the filesystem is currently 1.2PB of data on disk and once we have kept 90 days of changes its appearing as 1.6PB on tape. (or 307 LTO7 tapes)

We back this up using mmbackup and 8 threads and need to have a serial disk pool so that TSM will migrate the data in parallel, as we found that using a random disk pool mean that the migration only used one thread per a file-system.

The target is using IBM JD tapes hence why we can't export our copy pool to import the data, and create the replication. (We where previously/are currently keep two copies within the one tape library, but the tape library is currently full hence why we'v procured the second one, to provide us with the additional capacity)

After an initial issue with TSM running out of memory and needing to limit the DB2 memory (DBMemPercent 50) so the replication process could have enough, its now copying data to the new server if not a little slowly. (The source server has 128GB of memory)

The slow speed looks to be that TSM is only reading one tape on the source and therefore looks to be limited by the speed of one LTO7 tape drive. I can't see why TSM is only using one tape drive, we've given it two using "maxsess=2" and we've seen it use multiple drives if we replicate multiple filesystems (we only have 3 clients so this is a little difficult, (although there are more scale proxy nodes sending the data))

Is there a trick to get TSM to use multiple tape drives per a single filesystem for replication?

We are running IBM Spectrum Protect 8.1.8.

Thanks

Peter Childs
 
Hello , that would be a tough job and will consume a lot of resources, if you have the luxury of copy pool I would prefer that and ship those to the second servers and restore ,rather than replication
 
Hello , that would be a tough job and will consume a lot of resources, if you have the luxury of copy pool I would prefer that and ship those to the second servers and restore ,rather than replication

Don't really have that luxury given one is LTO7 and the other is JD. mind you we have a 10Gb/s network and we're already 25% done in the last month......... its just we'd like to get the task done before April.

Peter.
 
Back
Top