Target Storage Pool Cleanup

booman55

ADSM.ORG Senior Member
Joined
Feb 21, 2007
Messages
259
Reaction score
11
Points
0
PREDATAR Control23

I’ve got a customer that started doing Protect Storage Pool/Replication to another a secondary site and the WAN link wasn’t able to keep up and Protect Stgpool was running for days. They got into a pattern of running it until 6pm and then cancelling it. Replication hasn’t run in weeks. Suffice it to say, they ran out of space at target site.

At this point, I think they are better off starting new since most of the Oracle data will probably age off once replication and expiration is run.

What’s the best way to clean up the target server? Note, there are no local backups it’s just a target server.

  1. Protect stgpool purgedata=all
  2. Delete filespace on target server
  3. Nuke it and start a new instance
 
PREDATAR Control23

Not a fan of the bazooka approach.

Expiration, if it's running daily on the target server as it should would already taking care of expiring old data.

If they ran out of space on the target, sounds like they didn't size the target to match the source. They would just need to add space, then run protect to completion to copy all the data, that's the most efficient way to move data. Then run replication which will only need to copy the metadata.

If you blow everything away, you'll need to copy everything that's already been copied again.
 
PREDATAR Control23

Marclant, I agree with your assessment. The issue is Protect Stgpool won't finish and if it runs into backup window they're saying it affects backup performance. Being that Protect Stgpool isn't completing, Replication isn't running since running it would copy backup data and metadata.

Replication hasn't run for almost a month meaning that there's a lot of "expired" data taking up space. The target site actually has more disk assigned then the source site.

I'm leaning toward the delete filespaces for the Oracle backups. This should cleanup a ton of space. I'm guessing I'll then need to run protect stgpool with forcereconcile.

Once we have space cleared, I'm going to have them run protect stgpool around the clock until it finishes. They'll need to deal with a little pain with backups to get back in sync.
 
PREDATAR Control23

You can run protect and cancel it if you don't want to interfere with backup. However, protect does very little to interfere with backups as it's only doing DB transactions at the extent level, not at the inventory level like backups and replication.

I suspect the system is resource constraint if the backups and protect can't run at the same time. If you can't get protect to finish now because of system constraints, it's likely that you will have the same issue even if you clean-up the target server. Were this pair of servers built following the Blueprint: http://ibm.biz/IBMSpectrumProtectBlueprints ?

One to prevent protect to run during the backup window is to create an admin schedule to start towards the end of the backup window with a MAXRUNTime value so that it ends just before the next backup window.

Deleting the filespaces won't clean-up that much spaces, that will only delete extents for files replicated that are not referenced by other files. It will not delete extents that were protected, but not associated with a replicated file. The protect will take care of deleting old extents on the target as long as they are not referenced by other files on the target, which would be unlikely unless the target does ingest.

If you're going to start fresh, just break the replication between the 2 servers. Blow away the target instance, and create a new one. Can't be any fresher than that.
 
PREDATAR Control23

I assume by breaking replication you mean issue remove replserver command?
 
PREDATAR Control23

Yes.

However, if all you do is start with a fresh target without any other changes, I suspect that the same problem will occur. I recommend you match the servers to the Blueprint as much as possible. I realize you might have different hardware, but there are some things you can do similarly. Example:
- use SSD disk for the DB and active log
- have sufficient DB directories/LUNs
- sufficient storage pool directories
All the above is in Chapter 3, from Chapter 2, determine if the environment fits a small, medium or large blueprint.
 
PREDATAR Control23

Yes. The servers are matched and configured using IBM Blueprints. The issue started when customer was migrating clients too fast and didn't account for WAN throughput issue, so they couldn't complete daily Maintenance script which does Protect/Replication. If replication isn't running then expiration doesn't work effectively on target server.

Moving forward, we did some performance tuning which should help increase intersite throughput. We just need some space on target storage pool to get Protect Stgpool running again.
 
PREDATAR Control23

Ok, sounds like you might be on the right track then.

It's common in the beginning though for protect to take longer than usual to run. Sounds like the only issue right now is the space on the target. Just keep running protect until caught up, then start with replication. It's better to do those manually until steady state, then start with automatic schedules.
 
  • Like
Reactions: clj
Top