Chapter 7. Advanced maintenance, troubleshooting, and diagnostics 425
Logical drive bouncing
The constant transfer of logical drives back and forth between the two controllers can be
caused by a number of reasons. It can result in a severe performance degradation and should
be corrected. It is more likely to be a host configuration issue. In order to find evidence of
logical drive bouncing, we need to check the Major Event Log (MEL), looking for a constant
stream of non-critical event type 300D errors with a description of “Mode select for redundant
controller page 2C received”, as shown in Figure 7-66.
Figure 7-66 Mode Select 2C events
The DisableLunRebalance parameter
If the events appear exactly every 60 seconds, then it is possible that the
DisableLunRebalance parameter is set incorrectly in an RDAC/MPP clustered environment. If
a host in a cluster server configuration lost a physical path to a DS4000 or DS5000 storage
subsystem controller, the logical drives that are mapped to the cluster group will periodically
fail over and then fail back between cluster nodes until the failed path is restored. This
behavior is the result of the automatic logical drive failback feature of the RDAC multipath
driver. The cluster node with a failed path to a DS4000/DS5000 controller will issue a failover
command of all logical drives that were mapped to the cluster group to the controller that it
can access. After a programmed interval (normally 60 seconds), the nodes that did not have a
failed path will issue failback command for the logical drives because they can access the
logical drives both controllers, resulting in the cluster node with the failed path not being able
to access certain logical drives. This cluster node will then issue a failover command for all
logical drives, repeating the logical drives failover/failback cycle.
The workaround is to disable this automatic failback feature in all clustered configurations. In
Windows, change the DisableLunRebalance registry setting of the
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\rdacdisk\Parameters]
registry key from 0 to 3 in each cluster node. Then, reboot each node for the changes to take
effect.
To Next Page
Loading...
Need help?
General