[Linux-cluster] Throughput drops with VMware GFS2 cluster when using fence_scsi

[Greg Mortensen]

<Responding to myself for benefit of the list>

This looks like it was caused by the device mapping defaulting to a
round-robin path selection policy.  While I couldn't find any mention
of it in the RedHat cluster documentation, I did see some MSCS
postings[1] that said:

Round Robin can interfere with applications that use SCSI reservations
for sharing LUNs among VMs and thus is not supported with the use of
LUNs with MSCS.

so I changed the policy to "Most Recently Used" and was able to get
sustained writes of 40MB/s and sustained reads of 100MB/s down one
NIC.

  Regards,
    Greg

[1] http://en.community.dell.com/techcenter/storage/w/wiki/2671.aspx