VSP Storage Array - 6x MPB, 776GB cache
Microcode SVP 70-06-23/00
Microcode DKC 70-06-33-00/00
Storage Pool - HDT - 140TB capacity (208 disks)
3 tiers of disk - SSD / 10K SAS / 7.2K NL SAS
All parity groups configured with AES 256 bit encryption.
TrueCopy replication over dedicated circuit to remote VSP.
Brocade SAN comprising 5300 switches with 8Gb SFPs running FOS 7.3.1c
Dell PowerEdge R720 Server
SQL Server 2008
Microsoft Windows Server 2008 R2 Standard
2 x 8Gb HBA
Hi, we are seeing 100% utilization on one MPB on the VSP whilst a SQL DB rollback is being performed on the above SQL server. The LUN in question hosts the DB files and is assigned to the MPB.
This is affecting other systems whose LUNs are also owned by the same MPB with loss of service to underlying disk eg ESX Hosts.
Moving the busy LUN to another MPB causes the load to move to the new MPB and the old MPB utilization drops to more normal figures.
During the busy period of SQL rollback - approx 2.5hrs
The busy DB LUN has
approx 1200 IOPS,
Response times of 5ms-7ms
Read throughput of 10MB/s
Write throughput of 0-1MB/s
CLPR Max Write Pending Rate for the busy MPB flat-lines at 50%
Any suggestions how we can improve matters (other than buy more cache & MPBs) ?