Partner Path Misconfigured
Posted by richswain in Oct 13, 2009, under Tips and Tweaks
A packet is sent down the line and it comes to a fork in the channel, does it pick the right way? Only if you tell it.
With N series active/active configurations, you can see a lun from both controllers. This sometimes confusses people and they end up picking the wrong path to the hosting controller. The I/O that should be going directly to the primary controller is now going through the ‘partner’ and over the interconnect. This does increase the amount of workload each system has to put out as it handles the extra traffic.
Typically there are a few things to check when trying to diagnose the issue.
- Identify the LUNs being accessed through the partner node’s FCP Target port.
a. lun stats -o (LUN STATISTICS) - Identify the host initiators that are performing the I/O through the partner path
a. lun config_check -A (LUN CONFIG CHECK)
b. lun show -v (LUN CONFIGURATION)
c. igroup show -v (INITIATOR GROUPS) - Identify the primary storage controller FCP target ports available for access to the LUN.
a. fcp show cfmode (FCP CFMODE)
b. fcp show adapters (FCP TARGET ADAPTERS) - Verify the host initiator connectivity to primary FCP target ports and the host MPIO software configuration.
- Verify use of the partner path has ceased from both cluster nodes.
a. sysstat -b 1
Also make sure your MPIO settings are correct and not affecting the system path. All hosts should also be running the host utility kit provided by IBM.
With these techniques you should be able to clear up any mis-directed I/O. As with any correction, if you are not sure please contact IBM Support at 1.800.IBM.SERV or open a service request ticket at IBM Online Support Portal
October 25th, 2009 8:44 pm
Please also remember that VCB CAN cause multipathing issues, as VCB is only supported with 1 Path! meaning if you have luns on both controllers, you may need more than 1 VCB server.