How to simulate a disk failure - character device (ASM)

[Ken Englander]

We need to simulate a disk failure to check on flag in the minor number on /dev/async.  Running HP-UX 11i v3, Oracle 11gR2.  We cannot unplug the fiber connections because all the devices are on the same fabric.  We only want to fail a single LUN.

The bit settings are defined below.  We are setting it to 4.

Any suggestions would be welcome!

Thanks,

Ken

Disk I/O Hangs On HP-UX With ASM After Loss Of Storage [ID 393314.1]

Timeout setting of storage devices is configured too high or indefinite. When device is not accesible, ASM I/O requests hang till timeout expires. For example, /dev/async device file doesn't have support for timeouts if minor device number is 0x000000.

The minor device number is 0x000000. Here are the possible values for this:

0x000000 default
0x000001 enable immediate reporting
0x000002 flush the CPU cache after reads
0x000004 allow disks to timeout
0x000005 is a combination of 1 and 4
0x000007 is a combination of 1, 2 and 4

[Matti_Kurkela]

Modify SAN zoning to block communications to that LUN?

Deconfigure that particular LUN in the storage?

[Ken Englander]

Sorry, I should have mentioned I was looking for something easier than modifying the zoning.  That is secured so it is difficult to get done.

Any other suggestions?

[Matti_Kurkela]

Hmm, if you had ASM on a LV, you could do a "pvchange -a N" on the LUN, but you might not have such a configuration...

You said you're using 11iv3. How about:

scsimgr disable -D /dev/rdisk/diskNNN

 on the desired LUN?

You can restore it with the corresponding scsimgr enable command.

According to the scsimgr whitepaper, the system will automatically disable a LUN in the same way if an unrecoverable deferred disk error occurs, so this might actually be a relatively good simulation for your case.

The scsimgr whitepaper is available here:

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c01914672/c01914672.pdf

See pages 39-40.

[Ken Englander]

We are going to try that.  The scsimgr man page also mentions it and refers to the "block device".  Oracle ASM uses the character device, but I would imagine if the LUN is blocked, it does not matter which device is being used.  We will find out soon.

Thanks!