Copypool reclamation...long and confused

[JPFOLAN]

TSM 4.1.3 on NT4



I have a situation at present where my TSM server has been been running copypool reclamation all day for the last three days and only the ouput volume is mounted.



As I understood it when copypool reclamation takes place as the copypool should be offsite the coresponding tapepool volume is mounted and active data is copied to the new volume as the original volume is reclaimed.



I would therefore expect two tape volumes to be mounted for reclamation. This has been the case for the last 2 years of operation.



Can anyone explain how/why the copypool reclamation only requires the mount of the output volume?



During the reclamation process the backuppool apears to be active. It therefore looks like the backuppool is somehow involved in the reclamation process.



Any advice or experience of this issue would be appreciated.

 

[pbanghart]

Howdy...



When reclamation of a copy pool runs, it generates a new copy of the data in the associated storage pool(s)...this includes the disk storage as well as tape. I don't know how large of a disk storage pool you have, but perhaps that's the source that is being copied from, hence only one tape mounted, that being the new copy being generated...?



...just a thought...





-PB

 

[chad_small]

Sounds like single drive reclamation but in that case the source volume is mounted and data is read to the storage pool then the destination volume is mounted and the data is migrated off the diskpool back to tape. The problem you are having sounds fishy. I would try and clear it up by using MOVE DATA commands on any volume less than X percentage reclaimable. You can create a script that will process the tapes in order.



select volume_name from volumes where stgpool_name='<copy storage pool>' and pct_reclaim >=80 order by node_name



then use the following command



move data <volume name> stg=copypool wait=yes



This can all be handled by a good shell script.



Chad Small

[email protected]

 

[JPFOLAN]

Hi Chad,



Thank you for the info. I have added an additional observation.



It would appear to reclaim mountable volumes shortly after they have been checked out. I guess within an hour from checkout as I believe the reclamation process runs every hour.



This morning for example tape a and tape b were checked out. Within an hour of this reclamation started on tape a and tape b, tape c is mounted as the output volume and no other tapes are mounted. It appears to be reclaiming tape a and tape b from the diskpool as opposed to the tape pool. Could this be because the data on tapes a anb b is still in the diskpool and has not migrated to the tapepool?

 

[klebesco]

Your situation makes sense to me....



If you finished backups (disk primary pool to tape copy pool) and have not run migration as yet, the source will be the disk primary pool...



The reason your copy volumes are going into reclamation so quickly is because the are not full... If you only fill up 20% of a tape it will show filling and 80% reclaimable... therefore, if your reclaim threshold is 80 or under the tape will be reclaimed.... even though everything on it is good stuff.



Primary tape pools do not allow filling tapes to be reclaimed (because it is onsite and can hold more stuff)

The presumption with copy pool tapes is that its offsite and in reality there is no difference between a copy pool tape that once was full and now only 20% is good stuff... and the one above.



Perhaps you should consider blocking reclamation from happening except at your convenience.... update stgpool "copypoolname" poolreclaim=100



and make a script something like:

select volume_name, pct_reclaim, status from volumes where access = 'OFFSITE' order by pct_reclaim desc



decide how many tapes you want to reclaim and the when you want to start reclamation reissue to update stgpool command with a reclaim value you decided on.



a few minutes later (after the reclamation has started) again reissue the update stgpool with the value of 100.... the reclaim will continue to run until finished or something with a higher prioity kills it.



Why do this.... I don't know how many tape drives you have, but, if you allow the possibility of running rclamations while backups or migration is happening can possibly need more drives than you have and the reclamation could delay one of these more important processes, or you can have the same issue with reclamation/migration both wanting the same volume.



Just curious, is your primary tape pool collocated? If so and you have lots of nodes reclamation can seem to go on forever.



Good luck, Kurt

:grin:

 

[chad_small]

Question - You said the the tapes are in a mountable state. Are you using a MOVE DRMedia to checkout the volumes? If the state is mountable in TSM it will try to reclaim them. Make sure they are moved to a vault state and then I would not reclaim the tapes unless they are 80% reclamable, and a lot of times I use the move data command instead because then I can control the process.



chad small

[email protected]

 

[JPFOLAN]

Chad/Kurt,



Thank you for the updates and advice I think I understand what is happening here now.



Kurt,



I do not use colocation and have now setup to stop reclamation happening during the backup window.



hanks again guys

 

[finfrockg]

Previous post: The reason your copy volumes are going into reclamation so quickly is because the are not full... If you only fill up 20% of a tape it will show filling and 80% reclaimable... therefore, if your reclaim threshold is 80 or under the tape will be reclaimed.... even though everything on it is good stuff.



This statement is wrong. If your tape is 20% full and no data has expired on the tape, then it is 0% reclaimable. Make sure you do not set your reclaimation too low for the offsite copy. I set mine at 80% and it runs only every other day. I do have colocation active for the onsite tape pool (sequential access stg pool) and do not use colocation for the offsite copy. You'll appreciate the colocation with the onsite copy the first time you have to do a big restore. It dramatically reduces recovery times. Try to set reclamation as high as possible for your environment. Low reclamatin needlessly wears out tape drives and tape media. Two factors usually determine how high you can set reclamation:

the quantity of tapes available and the amount of "free" time your TSM server has to do reclamation. If you do not mind the extra offsite tapes, set the number higher.



Two tapes should always be mounted during reclamation- onsite volume and new offsite copy. If only one is mounted, it must be pulling the data from a diskpool.

 

[klebesco]

Dear finfrockq,



I beg to differ with you...



I was responding to:



"It would appear to reclaim mountable volumes shortly after they have been checked out. I guess within an hour from checkout as I believe the reclamation process runs every hour."



I would agree that mountable state cannot be reclaimed... BUT, as soon as a tape goes through checkout the state becomes vault, the access becomes offsite and lastly the reclaim percent jumps to 100 minus the pct. util. value ...and will become fodder for the next reclamation of that copy pool if the reclaim value is set to an equal or lower value.



I do like your funny face though! How do you do that?



Kurt

:lol: