Hi, all: I've got a question I've been wondering about, relating to how
foresightful TSM is in its' offsite reclamation. I'll spin an example:
TSM server 'SRV1' has three nodes
NODE1
NODE2
NODE3
It's got stgpools
PRIMARY : collocated . Volumes named Pxxx
COPY : non-collocated . Volumes named Cxxx
OFFSITE : non-collocated . Volumes named Oxxx
Daily incrementals happen on all nodes.
Tapes go offsite every day.
OK, given that scenario, we know that the vast majority of offsite tapes have
a thin slice of data from all of NODE1, NODE2, and NODE3.
--- Known scenario ---
So when an offsite volume O213 passes the reclamation threshold, the server
begins building a new offsite volume O501.
It mounts PRIMAY volume P001 (with node1's data on it) and makes an additional
copy of those files from NODE1 that appear on O213, and are still interesting
to the server.
Then it mounts PRIMARY volume P002, does the same for NODE2,
Then it mounts PRIMARY volume P003, does the same for NODE3.
Now, new offsite volume O501 is ready to leave on the next truck.
--- Unknown scenario ---
Now, what if O214, O215, and O216 all pass the reclamation threshold? We know
the server's going to start building O502. Say it picks O214 to begin with.
When P001 is mounted, will the reclamation copy files from -all- the
reclaimable offsite volumes, only from O214, or what?
I can see pseudocode something like:
- Pick an offsite volume to work on
- From that offsite volume, pick a first onsite volume to mount
- From that onsite volume, determine all files wanted for offsite
reclamation. Copy them.
This would be expensive in query time, but the alternative is a big-O N * M of
tape mounts, which makes me shudder. But it would certainly be simpler to
code.
|