SCSI Reserve/Release

[pyarelal101]

Guys,

I am running NBU 4.5FP_3GA we have a Shared Storage environment using SAN and we have 15 Media servers.The tape drives get reserved and never released we have to do a vmoprcmd to reset the tape drives. If anybody out there knows why this happens , or if there is a OS patch or a st.conf file or a Veritas patch to fix this, would like to hear from you .

Thanks

 

[comtec17]

Could give a little detail as to what your enviroment is:

What type of HBA's

What type of Drives

What type of Robot

Are you using ACSLS?

What type of switches?

 

[paddysan]

If you are using Netbackup on Windows Media Servers are you using the Veritas Tape Drivers ?
If you are using various OS, say including SUN you will have to update the drive definition in the st.conf to switch of scsi reserve/ release and pass control over to Netbackup.
If you provide a little more detail I might be able to help you.

 

[jrw1234]

Was there a fix for this, we are getting similar problems with SCSI reserves not being released.
NBU4.5mp5 - Emulex LP9K HBA - LTO2 Tape Drives via a SNC (Scsi Fibre Bridge).

 

[rdapilmoto]

We're also experiencing similar symptoms...tape drives are being reserved, and not released after backup jobs complete.

We're running NB 4.5 FP3 on Solaris 8
We have an IBM 3584 Tape Library with 24 LTO1 Drives
We use QLogic 2200F HBA's
McData 4500 Switches

Temporary fix is to manually reset the tape drives to release the media tapes. Help would be appreciated this way as well!

 

[jrw1234]

mr rdapilmoto
Thanks for that -we also use IBM librarys - any chance you dropping me a e mail so we can swop notes. jrw1 at hot mail dot com Thanks

 

[paddysan]

First thing to check is in the drive definition in the /kernel/drv/st.conf.
The default drive definitions do not disable scsi reserve/ release at the o/s.

tape-config-list=

"IBM ULTRIUM-TD1", "IBM Ultrium", "CLASS_3580",
"IBM ULT3580-TD1", "IBM 3580 Ultrium", "CLASS_3580";

CLASS_3580 = 1,0x24,0,0x41863d,2,0x00,0x01,0;

Also be careful with spacing/ tabs when adding this entry. The key part of this is in the CLASS_3580 string. This is obviously for IBM drives if you have HP LTO's e-mail veritas and they should be able to provide you with the correct settings to disable scsi reserve/ release.

Making this change means that netbackup now manages the reserve/ release for the drives, not the os. Also use the vmdareq -M <master server> command to view the reserve status of the units.

Also with the Emulex adapters with certain driver/ firmware versions robotic control commands can be lost down the pipe. So if you are experiencing drives getting stuck constantly and tape mounts/ umounts failing it might be useful to check the firmware first and then the drive compatability. My experience with this is normally, it at first it doesn;t work, try another combination as there are no hard and fast configurations. You can try the latest compatability matrix available from veritas for the supported versions as a starting point.

 

[jrw1234]

Paddysan

We are getting this issue in a Non-SSO enviroment as well.
I thought it might be a HBA issue thou, its also not explicit in the SSO notes for 4.5 that you need the st.conf entry. On the SSO kit we do have we are currently using the patched Solaris defaults

I might have to log a call with veritas to see what versions they have tested and marked as Supported.

 

[sgoldsmith]

You have the sam ething that we had, but in a 3.4 environment.

To release the drives without reseting them type the following:

to check for Reserved drives type the following:

vmdareq | grep RES

then compare the reserved drives with the actual drives in use in the silo (via ACSLS)

vmdareq -release -H media_server -D driveid

Or you can add the NO_TAPEALERT touch file to /usr/openv/volmgr/database. This will tell Netbackup that there is robot based cleaning occuring and to ignore the drive cleaning flag.

Hope this works... it does on 3.4 SP4 and greater!!

Simon Goldsmith
Storage Engineer


Simon Goldsmith
Storage Engineer

 

[Stumpr]

This may be a hold over from an older version of NB and SSO but ......

I place an entry for each server that shares the drive in the server's /usr/openv/volmgr/vm.conf file
DEVICE_HOST = server1
DEVICE_HOST = server2
DEVICE_HOST = server3 ... etc.