SQL Cluster Storage Changes.

This past weekend we had a client environment that needed to move from Microsoft iSCSI to Fibre for SAN connectivity. The reasons for this are numerous.

• Fibre connectivity has greater throughput.
• The Microsoft iSCSI initiator is junk for a clustered environment. ( when installing 2.07 the cluster nodes would intermittently lock up and lose resources, this is a known issue with 2.07. We updated to 2.08 and still had issues.)
• PowerPath on Fiber has proven to be solid numerous times.
• Less complexity.
• Greater reliability.

We ran into some issues though and since I didn't see any direct posts on the web or on TechNet. So I will detail the issue and the solution here.

Specs:

Servers HP DL380G5; Windows 2003 R2 Enterprise 32-bit, SQL Enterprise 32-bit. Current cluster configuration is Active/Passive. Initial storage configuration is iSCSI. Moving to Fibre HBAs.

Issue:

After installing a Fibre HBA Dual port card and zoning it to the storage. We uninstalled the MS iSCSI initiator and disabled the iSCSI NIC ports on the server. Upon rebooting we attempted a full failover. The cluster group would fail over with the Quorum but the SQL instance would fail over the IP resources and the T log drive then hang. The result is a full failover is not possible, which in an environment that HAS to be online is not a good thing. The event ID would reference not being able to flush the transaction log.

Solution:

After reboots and trying to see if we missed anything these are the steps that we followed to resolve the issue:

• Uninstall from windows the NICs used for iSCSI.
• Remove the physical adaptor used for iSCSI from the server.
• Reboot.

Once these steps were completed a full failover was again possible.

If you have any questions/concerns or if you are having this issue yourself please feel free to comment. We will do our best to assist you.

CC