Originally posted by: Alexs
I am Linux/unix admin, who is looking after some Linux Oracle rac clusters. The storage is a HDS system (not sure what it is ). I was wondering do HDS publish any best practice configs for setting up multipathing on a Linxu (RHEL 5.5) box. Previously I have worked on site with EVAs and XP and I have relieve upon hp's spock for whitepaper and best practices.
We have had some issues with multipathing not being setup properly and seeing pathing issues. I believe (as I don't have access to the HDS interface), that a lun changed its presentation from one controller to another and this cause problems with a couple of the Linux boxs.
I came across this in my googling
This is what I could find about this configuration variables
This setting controls the failback scan task behavior of rebalancing the LUNs or putting the LUNs back on their preferred owning controller, when failed paths are determined to be good and usable again. This parameter can take on values from 0 through 3.
0 - means LUN rebalance is enabled for both AVT and non-AVT modes (default)
1 - means LUN rebalance is disabled for AVT mode and enabled for non-AVT mode
2 - means LUN rebalance is enabled for AVT mode and disabled for non-AVT mode
3 - means LUN rebalance is disabled for both AVT and non-AVT modes
DisableLUNRebalance deals with redistributing volumes to their preferred controllers, when possible. For clustering DisableLUNRebalance is recommended to be set to 3.
This setting provides an indication of how the driver should behave when it finds a LUN owned by a controller that the driver does not have a working path for. Zero means leave ownership as is and fail the I/O; one means fail the LUNs back to the controller, which does have a path.
FailbackToCurrentAllowed only comes into effect when a failover decision is being made, and is a legacy carryover. The value is always set to 1 on Windows/Linux platforms.