HSM priority and mountpoints

drq

ADSM.ORG Member
Joined
Nov 8, 2008
Messages
6
Reaction score
0
Points
0
PREDATAR Control23

All,

I recently started to share a library between two TSM instances. I thought I calculated the scheduling needs/drives, scratch tapes etc. One thing I failed to plan for is now proving to be quiet a pain.

09/21/09 09:55:46 ANR1440I All drives in use. Process 13 being preempted by
higher priority operation. (SESSION: 23199)
09/21/09 09:55:47 ANR0409I Session 23199 ended for server MCTSM02
(Linux/x86_64). (SESSION: 23199)
09/21/09 09:55:48 ANR1020W Migration process 13 terminated for storage pool
DISKPOOL - process canceled. (SESSION: 22342, PROCESS:
13) I only have one HSM enabled client (I very thankful for this) associated with one of the TSM servers. I've configured the node on the server side to only use 1 mount point and I believe I set a very low priority on the nodes schedule (prob doesn't apply to the actual job). It would appear that the HSM client is seemingly mounting tapes etc virtually all the time. I'll admit that I'm not sure how it is supposed to operate but I do know that it seems to want to mount tapes throughout the day and it's schedule starts at midnight running once per day. The node is scheduled with this option
-[SIZE=-1]-subdir=yes -postschedulecmd="dsmautomig"[/SIZE]

1. TSM prioritizes restores so i suppose HSM falls into this category. Anyone know of a way I could make HSM be the lowest priority?

2. I'd like HSM to migrate files after the backup - I don't believe anyone is recalling files. What could cause it to run like this. I believe the HSM client v is 5.5.2

Thanks
 
PREDATAR Control23

Hi,

I may be able to help you. I have been working with hsm for aix for a while now. from your post I don't have much to go on at this point but first please post the output of the command dsmmigfs query this will show what the current settings for high threshold, low threshold and other setting are for the file systems you want to HSM.

Since you state that hsm is recalling from tape all the time how many drives do you have?
There is a server option NOPREEMPT that is a Global option but would stop any preemptions.
I would NOT reccommend setting this because scheds for Db backup, reclaimation, migration, etc will not be able to get control of the drives.

Generally you want to try to keep as much of the files on the disks as you can so that the users of the file systems dont need to retrieve from tape but that can be adjusted based on the size of the file system and how many new files are added, the number of files needed to be retrieved due to a user needs. basically Hsm make a fixed file system appear much much larger than it truly is. I have a customer with a 500GB filesystem that holds 87TB according to HSM. The files are replaced with 4kb stub files so as long as I dont fill the file system up with 4k stubs

Bob...
 
PREDATAR Control23

I don't have client side access to the HSM node - Something that is an issue. Windows admins want me to handle everything, aix,linux and solaris admins don't even give me a local login.

Can I tell what operation HSM is performing from the server only side?

The way this is setup makes me believe that no reclaims should happen. HSM services a clinical radiology application. The application writes to a online archive ~7TB. When a new image file is copied to the onsite archive it is also copied to HSM via a CIFS shared mountpoint on the HSM server. Nothing is ever deleted from the archive to my knowledge. No one logs into the HSM server and recalls images (to my knowledge) This setup was done to leverage the low cost of HSM running on AIX.

Thanks,

Bob
 
PREDATAR Control23

Hello again,
You should be able to see what hsm is doing by moniitoring perhaps in real time what sessions are connected to the TSM server. If the HSM client has a session open you can
q sess f=d and see if the hsm session has a tape open for read or write operation.

It may help to know more about your tsm enviroment. library, # of drives, size of the file system you have hsm'ed. With HSM if you are near the upper limit of the file system size that may explain why there are so many tape operations.

You could also ask the windows team to run the commands "dsmdf" and "dsmmigfs query" for you and send the output back ta you. In addition I use scripts with dsmmigrate -P /filesystem to do the premigration so that when the file system thresholds are exceeded then the files have already been migrated and all that needs to be done is to replace the files with the stub pointers. I hope you are using a policy for this client that requires backup before migration?

Bob...
 
Top