Pillar FalconStor System
 
Pillar Axiom 600 Storage System
 

PLA-Support>Sun Systems>DISK>Axiom>SN-DK: Ax600
 

Applies to:

Symptoms

FalconStor IPStor Disabling Storage Paths
IPstore hosts are getting below error messages while Axiom is reporting Non-Optimized Access to LUNs:
29/00 : power-on, Reset, Target Reset, or LUN Reset occurred
2A/06: Asymmetric Access State Changed

Changes

IPstore is upgraded from 6.x to 7.x (below 7.50)

Cause

Axiom sends the queue-full sense code for a number of reasons (busy, command timeout, out of resources), but the meaning is simple: Please retry the command again. This is not an error condition.

The real heart of the issue here is that Queue Full/Task Set Full is not a condition that should induce a path failover on the IPStor side. That has no bearing on the status of the path and trying on a different one probably will make no difference. It can be destructive if non-optimized accesses are being performed which they are here.

The error scenario is similar to below:

1) Everything is fine while load is normal.
2) Load goes up. Axiom gets overloaded and commands start to come back busy or timed out.
3) SAN returns Status=0x28 Task Set Full to IPStor.
4) IPStor reacts by changing paths and retrying
5) Command may succeed or it may get returned again with TSF. This causes further path failover.
6) NOAs start due to loaded Axiom and the unnecessary path failovers
7) Performance further drops due to NOAs and lun ownership changes

 

Solution

Oracle Pillar Axiom engineering has recommended that implement a path failover timeout of 30 seconds, over the current 10 seconds to FalconStor engineering.
We do not have much thing to do on this issue.

References

This solution has no attachment