TSM Library Manager AIX OS settings for VTL drives

Nicke

ADSM.ORG Senior Member
Joined
Mar 17, 2005
Messages
89
Reaction score
6
Points
0
Location
Sweden
PREDATAR Control23

I have a TSM environment with a libray manager and 100 Storage agents running mainly in AIX.
The Library Manager has also a normal TSM server role with about 400 clients.
The backup environment consists of 2 IBM ProtecTIER clusters and a toltal of 4 virtual libraries.
We have a total of 384 virtual tape drives accessed by 8 HBA ports in the Library Manager and among the Storage Agents.

The Library Manager runs in AIX (7.1) and on a Full partition (LPAR with no virtual devices).

...

Periodically we experience ANR8311E and ANR0523W errors which lead to DB2 API session failures.

We can't determine the root cause and no AIX errpt errors show and also no problems with the SAN fabrics or any alerts from IBM ProtecTIER.

I'm looing at the HBA port settings for both the Libarary Manager and all Storage Agents now.

-> My question is really what is the current HBA setting recommendation for VTLtape when using IBM Power 7 (8Gb PCI Express Dual Port FC Adapter), TSM 6.4 and AIX 7.1 ?

Should we use "fc_err_recov" = delayed_fail or fast_fail and whats the recomended dyntrk setting?


Thanks for some input :)
 
PREDATAR Control23

Without looking at the fiber traffic to be 100%, I am guessing that 8 FC connections is NOT enough for 384 Virtual Tape Drives.

This gives you a ratio of 48 tape drives per FC. I suggest increasing the number of FC so ration is at least 25 (or less) Virtual Tape drives per FC.

Also, the errors mentioned point to access errors and not specifically to FC channel overrun errors. Have you looked at the resource utilization values and/or maxnummp settings? You may have over provisioned for the number of Virtual Tape Drives available vis-a-vis the nodes/storage agents attempting a backup.
 
PREDATAR Control23

Without looking at the fiber traffic to be 100%, I am guessing that 8 FC connections is NOT enough for 384 Virtual Tape Drives.

This gives you a ratio of 48 tape drives per FC. I suggest increasing the number of FC so ration is at least 25 (or less) Virtual Tape drives per FC.

Also, the errors mentioned point to access errors and not specifically to FC channel overrun errors. Have you looked at the resource utilization values and/or maxnummp settings? You may have over provisioned for the number of Virtual Tape Drives available vis-a-vis the nodes/storage agents attempting a backup.
...
Thanks for your input MoonBuddy

We have a couple of IBM tsm pmr's open (since 3 weeks) for these vtl tape access problems but still no answer about the root cause...

All tsm options are checked by both myself and IBM Tsm support but everything looks ok.

And this is actualy an AIX question but it's also a TSM Tape question that I hope someone has answer for ?

Should we use AIX "fc_err_recov" = delayed_fail or fast_fail for the hba fscsi# and whats the recommended dyntrk setting?

The only IBM Aix info I can see regarding recommended hba settings is for Disk and NO actual info regarding hba settings specifically for TAPE
 
PREDATAR Control23

Use the recommended setting for the disk and work your way from there. Truly, VTL are disk operations and should be treated as such.
 
Top