PP HP-UX , No Pseudo

Hi. The Powerpath driver inserts itself into the I/O stack at the SCSI driver level (in some sense assuming its identity).

Essentially, all I/O and Ioctl operations to the SCSI driver first pass through PowerPath.
For those devices PowerPath is managing, it may load-balance, i.e. redirect the I/O or Ioctl to the SCSI driver on an alternate path (to the same physical LUN). Otherwise, the I/O or Ioctl is simply passed-through to the SCSI driver.

Hope this helps.

-Brion

so powerpath going install some driver in hp-ux kernel , those driver or module going check evey io application layer in os will send for scsi devices and based on that check if this is a san device and do failover and load balancing
AM I right ?????????????????!!!!!



Symmetrix ID=XXXXXXXXXXX
Logical device ID=0071
state=alive; policy=SymmOpt; priority=0; queued-IOs=0
==============================================================================
---------------- Host --------------- - Stor - -- I/O Path - -- Stats ---
### HW Path I/O Paths Interf. Mode State Q-IOs Errors
==============================================================================
0 0/0/4/1/0/4/0.118.0.1.0.8.6 c0t8d6 FA 8aB active alive 0 0
2 0/0/4/1/0/4/1.166.0.1.0.8.6 c2t8d6 FA 9aB active alive 0 0

in this case if i assigned c0t8d6 as raw device for oracle and the path to FA8ab failed will powerpath failover to c2t8d6


waiting u r answer

fail over should work regardless if the devce is used as raw or not.

powermt watch can show how the load is currently getting distributed across the paths. Lok at IO/Sec

110 0/0/8/1/0.98.33.19.0 optimal 57 0 376 0 0
114 0/0/10/1/0.97.33.19.0 optimal 57 0 401 0 0
118 1/0/8/1/0.97.33.19.0 optimal 57 0 404 0 0
122 1/0/10/1/0.98.33.19.0 optimal 57 0 376 0 0

As others have said, on hpux, PP will intercept any I/O going to ctd devices, whether raw or cooked, even if Oracle is defined using just one of the two paths. This is akin to creating a VG with just one path and not the alternate path. PP will handle the balancing and the failover even if the path defined in the VG fails.

You can verify the load balancing by running "powermt display every=10"

Keep an eye on the IO/sec column for activity across all paths.