ISCSI Failover time

[jeffdorn]

Just implemented an additional LeftHand SAN to vSphere infrastructure. Our current environment is Fiber, so this is my first experience using ISCSI. I've set up 2 ISCSI adapters to load balance with Round Robin per VMware's documentation (http://www.vmware.com/pdf/vsphere4/r40/vsp_40_iscsi_san_cfg.pdf). Everything seems to work properly except when I test for a failover. While I'm in a windows host copying data to the SAN, I manually will shut down a port on the Cisco switch. Windows will take 40 seconds before data starts copying on the other path.

Is there something I missed? 40 seconds seems like a long time for the failover to occur.

[FranckRookie]

Hi Jeff,

I couldn't find any reference to an "official value" for timeout on iSCSI storage (maybe someone can help). But 40 seconds doesn't hurt me. Some storage providers even encourage to increase default disk timeout detection inside guest OS when storage is on Ethernet. This way, no failure event is generated in the VM when the failover arises and the ESX has the time to switch to another path.

Regards

Franck

[Phokay]

As Franck's rightly said, you will want to increase the iSCSI time out on Windows registry to avoid apps error during the failover. Normally most iSCSI storage vendors recommend to increase iSCSI timeout value to 60 seconds. The other recommendation from VMware is to set "Disk.Device.Reset" to 0 and "Disk.LUN.Reset" to 1. By default they are both 1 (enabled). That configuration is under advanced configuration->Disk.