Does behavior of RHEL native MPIO (DM) vary for 11gR2 vs. 11gR1?
Hi,
I have two Oracle RAC instances configured - 11gR2 with RHEL 6.4 nodes and 11gR1 with RHEL 5.8 nodes. All nodes are using the Linux native Device-Mapper multipath IO tool.
I have ASM configured on external SAN devices. There are two paths to get to any given ASM data disk dirve.
However, when I simulate some 15000 transactions per minute to the separate RAC instances, the transaction IO gets distributed between the two paths going to every data device on the RHEL 5.8 instance, while on 6.4, it only picks one of the paths to run all the transactions and there is zero traffic on the other path. Thus there is no load distribution across the links. Is that expected and working as per design?