Starwind and Equallogic co-exist configuration
Posted: Sat Dec 31, 2016 9:17 pm
Hey guys,
We recently upgraded to a two-node vSphere 6 solution with Starwind running as a VM on each node. Each node as 7 TB of usable SAS SSD local storage. Our "old" hosts were using shared Equallogic SAN storage. We would like to continue to use the Equallogic SANs for seldom-used VMs and backup purposes.
Each host is connected via back-to-back 10 Gb ports. Two ports are for iSCSI and two ports are for Sync between the Starwind VMs. Each of the iSCSI ports also have a VMKernel adapter. The iSCSI ports are on separate vSwitches and different subnets, per the recommendation of Starwind. None of the VMKernel ports are bound to the VMWare Software iSCSI initiator (again per the recommendation of Starwind.)
Equallogic has a different approach. Each interface on the Equallogic boxes must be on the same subnet. They also use a "group" address that is used as the iSCSI target address in vSphere. We have 4 1Gb ports in each host that are dedicated to the EQL iSCSI. Each of these ports have an associated VMKernel adapter. The difference is Equallogic wants these ports on the same vSwitch. They also want these ports bound to the VMWare Software iSCSI Initiator.
This initially appears to work as designed. We started by configuring Starwind and we went through the very thorough session with Starwind support to verify we had everything configured properly. All looked great. We then configured the 1Gb ports for the Equallogic SANs, including binding them to the Software iSCSI Initiator. This all went well and everything worked. We could see multiple paths for the storage, etc.
However, if we restart a host, the host will either not connect to the Starwind volumes at all, or will only connect on one of the hosts; not both. The hosts connect to the Equallogic volumes consistently. Rescanning the HBA does not make a difference. (Note: we have the PowerCLI scripts configured in Starwind to tell the hosts to rescan when necessary.)
Here's the final interesting point: While the host is running, if I remove the iSCSI VMKernel bindings of the 1 Gb ports for the Equallogic SAN, then immediately the host will connect to the Starwind host. I can then re-add the Equallogic VMKernel adapters and things continue to work - until the next reboot.
I opened a ticket with Starwind, but since this doesn't appear to be a Starwind specific issue, they couldn't help further. I hoping someone has seen this before and has an idea to help.
Thanks
We recently upgraded to a two-node vSphere 6 solution with Starwind running as a VM on each node. Each node as 7 TB of usable SAS SSD local storage. Our "old" hosts were using shared Equallogic SAN storage. We would like to continue to use the Equallogic SANs for seldom-used VMs and backup purposes.
Each host is connected via back-to-back 10 Gb ports. Two ports are for iSCSI and two ports are for Sync between the Starwind VMs. Each of the iSCSI ports also have a VMKernel adapter. The iSCSI ports are on separate vSwitches and different subnets, per the recommendation of Starwind. None of the VMKernel ports are bound to the VMWare Software iSCSI initiator (again per the recommendation of Starwind.)
Equallogic has a different approach. Each interface on the Equallogic boxes must be on the same subnet. They also use a "group" address that is used as the iSCSI target address in vSphere. We have 4 1Gb ports in each host that are dedicated to the EQL iSCSI. Each of these ports have an associated VMKernel adapter. The difference is Equallogic wants these ports on the same vSwitch. They also want these ports bound to the VMWare Software iSCSI Initiator.
This initially appears to work as designed. We started by configuring Starwind and we went through the very thorough session with Starwind support to verify we had everything configured properly. All looked great. We then configured the 1Gb ports for the Equallogic SANs, including binding them to the Software iSCSI Initiator. This all went well and everything worked. We could see multiple paths for the storage, etc.
However, if we restart a host, the host will either not connect to the Starwind volumes at all, or will only connect on one of the hosts; not both. The hosts connect to the Equallogic volumes consistently. Rescanning the HBA does not make a difference. (Note: we have the PowerCLI scripts configured in Starwind to tell the hosts to rescan when necessary.)
Here's the final interesting point: While the host is running, if I remove the iSCSI VMKernel bindings of the 1 Gb ports for the Equallogic SAN, then immediately the host will connect to the Starwind host. I can then re-add the Equallogic VMKernel adapters and things continue to work - until the next reboot.
I opened a ticket with Starwind, but since this doesn't appear to be a Starwind specific issue, they couldn't help further. I hoping someone has seen this before and has an idea to help.
Thanks