Faulty symptom:When customer changed the owning controller, also the number of optimal paths changes on VMware ESXi. When LUN's have owning controller B, VMware see 2 optimized paths (which is correct, only the two paths to the owning/working controller are optimized paths). When LUN's have owning controller A, VMware see 4 optimized paths (which is incorrect, this should be 2).
Version information: VMware ESXi 6.0.0 Update 2 + 5500V3 V300R003C20
Network topology: Typical dual-switch network, each controller has two paths to host, 4 paths in total.
1. Check the mapping view and port group setting, everything is normal.
2. Chech the event log and found customer set the ALUA mode as "Common ALUA"
1250 2017-03-23 14:33:37 0x200F02440005 Event Informational succeeded in modifying the path performance type of initiator (type iSCSI, identifier iqn.1998-01.com.vmw) to optimized.
1249 2017-03-23 14:33:37 0x200F02440004 Event Informational succeeded in modifying the failover mode of initiator (type iSCSI, identifier iqn.1998-01.com.vmw) to common_alua.
3. We confirmed that customer doesn't have HyperMetro license. In this case, we confirm the inappropriate setting cause this issue.
1. In earlier version of V300R003C20, we only have ALUA mode, enable or disable. And we enabled ALUA on FC initiator in default, only need to enable ALUA on iSCSI initiator when needed.
2. We add three optional ALUA mode from V300R003C20, because we begin to support third part multipath for HyperMetro solution. In earlier version, HyperMetro only supported by Huawei UltraPath.
3. Common mode is designed for HyperMetro solution. In this mode, we only report preferred path(port group) for working controller. But, in the Early-version mode, we report preferred and non-preferred path on both of the controllers. This is the difference. We opened case to VMware support before, they replied us that they need to report path on all the controllers. In the non-HyperMetro scenario with Common ALUA, we only report a part of the path, they can’t support, thus the result is not expected.
Change ALUA mode from "Common ALUA" to "Early-version ALUA".
1, After change ALUA setting on storage side, please reset the host to make the change into effective.