Emptying stale data from and active data pool

[foobar2devnull]

Hi,

I have an active data pool which seems to hold data from servers which should not be found there. The fault is entirely mine but I'd now like to remove the data.

How can I clean the active disk pool of data from "SERVERNAME"?

I did a

q content /tsmfile/000015BE.BFS

to find what servers were still using the active data pool. Then a

q occ SERVERNAME

does indeed show that the server does use the active data pool (FSID 3), but a

q fil SERVERNAME

does not split storage pools but sorts by partition, so I have no way of just saying, delete files in the active disk pool for SERVERNAME.

Thank you for any help!

 

[moon-buddy]

Please clarify:

Do you still need the ENTIRE data for servername, or just want to get rid of the data belonging to servername on /tsmfile/000015BE.BFS?

 

[sjl]

The best option I can think of would be to mark all non-empty active pool volumes read only, then do a MOVE NODEDATA SERVERNAME FROMSTG=ACTIVEPOOLNAME, to consolidate that server's data onto empty volumes. Once the MOVE NODEDATA completes, use QUERY NODEDATA SERVERNAME STG=ACTIVEPOOLNAME to get a list of volumes holding the data for that particular server, and finish with DELETE VOLUME DISCARDDATA=YES for each such volume.

You can lather, rinse, and repeat for each server you need to push out of the active pool. Or you can run the MOVE NODEDATAs in sequence, and then finish off with the DELETE VOLUMEs.

Be careful not to run a COPY ACTIVEDATA while the MOVE NODEDATA process(es) is/are running, or you may have to re-run it after the deletion processes complete.

You could also do a QUERY NODEDATA before the MOVE NODEDATA to get a list of volumes, then QUERY NODEDATA * VOL=VOLNAME to check whether the data in that volume is exclusively servers you want to nuke from the pool; if it is, you can run the DELETE VOLUME for that volume before the MOVE NODEDATA, to cut down on the data movement. Depends on how much data you need to clean out.

Be very, very careful when running DELETE VOLUME; check, double check, and triple check before saying "yes, go ahead and delete it", lest you delete something important by accident.

 

[foobar2devnull]

I like how answers are coming from all corners of the world!

sjl got what I am trying to do. Flush one or more servers from the active data pool without loosing data from these servers on other storage pools. Who knows why I wan't able to describe it this way in the first place?

Thank you both for taking the time to answer. Great community. I hope I can return the favour sometime.

 

[sjl]

I like how answers are coming from all corners of the world!

All together now: "It's a small world after all, it's a small world after all ..."

One other thing: if you don't mind nuking data that should be kept in the active pool, you can just do the QUERY NODEDATA SERVERNAME command, then DELETE VOLUME on all of those volumes. The next COPY ACTIVEDATA will reinstate the extra data that was deleted, so the extra deletions will be a transient thing. It's a tradeoff: more data movement before the deletion, or after it? You can also do the QUERY NODEDATA * VOL=volume to figure out whether it's better to just nuke that volume, or leave it be and pull out the server's data using MOVE NODEDATA.

It comes down to which tradeoffs you want to make, like so much in IT in general.

 

[foobar2devnull]

I'm still learning a lot and your posts have been very helpful. The ADP was supposed to be migrated to another in an attempt to define some kind of SLA. In the end, a greater project is going to take over so I might as well re-use the ADP. I evaluated your commands and the data on the ADP and decided I could wipe all files in the ADP and define new servers to use it once cleaned.

Thanks again for your help. Very instructive.