Losing iSCSI connection during failover

Article ID: 134

Last updated: 18 Feb, 2009

Symptom:

I am testing iSCSI failover with Virtualiron. Virtualiron needs to assign iSCSI initiator port on the network node. Also, it needs iSCSI target's IP address.

First, I entered DSS's virtual IP address in the target's address. At this configuration, whenever changeover happens, Virtualiron loses the connection. And unless I restart the system which has the iSCSI initiator port, it does not discover the changed over system.

For example:

The iSCSI disk based on virtual IP of DSS is on line at VI center. I manually failover from system 1 to system 2. The iSCSI disk becomes offline. If I restart the system (node) which has the connecting port (initiator port) in, it discovers the virtual IP and the iSCSI disk becomes online again.

I manually sync volume and fail back from system 2 to system 1. The iSCSI disk becomes offline. If I restart the system (node) which has the connecting port (initiator port) in, it discovers the virtual IP and the iSCSI disk becomes online again.

So, I entered both virtual IP and physical IP address of 2 DSS systems, which is 192.168.5.230 for VIP, 192.168.5.221 for system1 and 192.168.5.222 for system 2.

When the change over happens VI loses 192.168.5.230 and the volume connected with it. However, the simple iSCSI volume in the system1 kept the ONLINE status.

And Virtual iron never discovers 192.168.5.220.

Problem:

Both Virtual Iron and VMware has this problem with Lun 0 and Multiple Targets with the same Lun Number.

Here is the link for VMware with this issue

http://communities.vmware.com/message/701181#701181

Solution:

Target 0 (Target name must be the same on the Secondary as well)

LUN 1 (try without using 0) SCSI ID (must be the same on Primary and Secondary)

LUN 2 SCSI ID (must be the same on Primary and Secondary)

Target 1

LUN 3 SCSI ID (must be the same on Primary and Secondary)

LUN 4 SCSI ID (must be the same on Primary and Secondary)