A preferred path critical event can be caused
by several
activities. You can resolve the critical event by trying a series
of corrective actions until one of them works.
About this task
Several normal
activities can generate a preferred path
critical event on the disk controllers in a TS7650 Appliance or TS7650G.
- Restarting
a server in a clustered configuration
- Failover between servers
in a clustered configuration
Note: Unwanted
fencing is a known problem in a clustered environment. Fencing is
the disconnection of a node from the shared storage of the cluster.
Fencing cuts off I/O from shared storage to ensure data integrity.
If you experience unwanted fencing and cannot stop it, contact your
next level of IBM support.
- Issuing a scan operation
from ProtecTIER® Manager
- Generating
a diagnostic summary
Each scan event causes the host bus
adapter (HBA) initiators
to query both the A controller and the B controller of the disk controllers.
This results in the Automatic Volume Transfer/Automatic Data Transfer
(AVT/ADT) being signaled and the logical units (LUNs) being moved
to those controllers for a brief moment. In a properly configured
environment, after the devices are actually used by the server, the
logical drives move back to the preferred path. In a configuration
that does not access every LUN (idle node, archive data, and so on)
those LUNs might not be accessed on a sufficiently regular basis.
Procedure
To resolve a preferred path critical event, try
the following
corrective actions in order until one of them is successful.
- Within Storage Manager, select .
Important: Do not select Change
Preferred Path to resolve the temporary messages. Only
perform this when you are correcting a misconfiguration.
- If the drives immediately or continually revert
to the "Logical
Drive Not on Preferred Path" state, type the command multipath at
the CLI prompt of the affected node. This normally
realigns the preferred paths.
- If the multipath command
does
not resolve the problem, type multipath -v2.
- If the multipath -v2 command
does not resolve the problem, use Storage Manager to modify the AVT
Default Host Type to LNXCLUSTER.
Important: If you are prompted to
change the password, do not change it.
To change the Default Host Type:
- Select .
- In the New
host type field, type LNXCLUSTER and
click OK. If you are prompted for a password,
enter ibm2serv.
- Repeat substeps 4.a. and 4.b. for all Storage
Subsystems.
- If the problem persists, open a
service call to either have the drives reconfigured or diagnose failing
hardware components.