Details
To ensure the IBM DS3200 SAS storage array provides proper path failover during heavy I/O in a VMware environment using ESX Server 3.5 Update 1, you must reconfigure the Remove Device Missing Delay (RDMD) timer value to 10 seconds and the I/O Device Missing Delay (IODMD) timer value to 0 seconds in the BIOS for each SAS HBA. Failure to do so might cause I/O timeouts in the event of a SAS HBA failure or cable disconnect. The following error messages might be displayed in /var/log/vmkernel:
Jan 29 10:59:29 stgbladeh101 vmkernel: 0:21:06:58.523 cpu1:1062)<4>mptscsih: ioc1: Attempting task abort on id=0! (sc=0x3e229200)
Jan 29 10:59:29 stgbladeh101 vmkernel: 0:21:06:58.523 cpu1:1062)<4>mptscsih: ioc1: Abort ctx2abort=e00a1
Jan 29 10:59:29 stgbladeh101 vmkernel: 0:21:06:58.537 cpu1:1062)LinSCSI: 3201: Abort failed for cmd with serial=31623, status=bad0001, retval=bad0001
Solution
1. Reboot the ESX Server host.
2. When prompted during POST, press Ctrl+C to start the LSI Corp (SAS) Configuration Utility.
Note: If you experience PCI resource allocation errors that prohibit the configuration utility from initializing, refer to KB article1004486.
3. Select the first SAS adapter, and choose Advanced Adapter Properties.
4. Under Adapter Timing Properties, change the Remove Device Missing Delay to 10 seconds.
5. Change the I/O Device Missing Delay to 0 seconds.
6. Repeat steps 3 and 4 for all SAS HBAs in the host systems.
Based on VMware KB 1004313