This post is more than 5 years old
129 Posts
0
48877
January 17th, 2013 23:00
Storage Heartbeat Vmkernal port for Equalogic PS array
Dell have just installed a new EQL PS SAN and configured it for our 5 x ESXi 5.0U1 hosts. Each host has 2 dedicated 10GB NICS on a dedicated vswitch for the iSCSi connections to the SAN. MPIO is via the add-on Dell MPIO driver.
They are configured in a 1:1 way to 2 VMK ports iSCSi1 ands iSCSi2 as per usual. However since Dell left I have come across a Dell/EQL white paper (TR1075-Configure-iSCSI-Connectivity-with-VMware-vSphere-5.pdf) which says that a separate vmk port should be configured to maintain "pingability" of the iscsi ports during failure/failover. This port should have an ip address on the iSCSi subnet and the 2 10GB nics should be active in the team but not be bound to the sopftware iSCSi adapter. It should also be the lowest vmk port no.
The following posts in various net forums seem to suggest this is vital:-
http://communities.vmware.com/message/1877959
http://www.linkedin.com/groups/New-guidelines-VMware-iSCSI-65187.S.89647775
http://sostech.wordpress.com/2012/01/10/iscsi-multipath-for-equallogic-round-3/
If it is there it is easy to create the additional vmk but how would I make it the lowest port without deleting the existing ports and interupting SAN access.
Regards,
Gerry
0 events found
Origin3k
6 Operator
•
2.4K Posts
•
12.3K Points
1
January 17th, 2013 23:00
It have to be the lowest VMK # of all your iSCSI ports, so you have to remove the 2 existing VMKs first and start from scratch. Reason is that vSphere use the lowest one to try the access the array when discovering/scaning the volumes on the EQL. If this pNic, Cable, pSwitch goes down and you press the Rescan button within vSphere you get into trouble.
With vSphere 5.1 this is not necessary any more so you may consider and upgrade to the latest version.
Regards,
Joerg
Dev Mgr
6 Operator
•
9.3K Posts
0
January 18th, 2013 07:00
I've been told that as of ESXi 5.1.0 the storage heartbeat is no longer needed due to a change in how ESXi responds to pings or so.
gerrybarnett
129 Posts
0
January 18th, 2013 07:00
Thanks for the quick response. Rather than delete both existing iscsi vmk ports on the switch, would unbinding and deleting the existing lowest port i.e ISCSI1/vmk1 , add the heartbeat port as vmk1, then recreating ISCSI and rebinding it work? That would seem to maintain SAN access throughout
swoods79
6 Posts
0
January 28th, 2013 07:00
I currently have two PE R910's connected via stacked PC 6224's to an EQL PS4100X.
My vSwitch for iSCSI was originally configured by Dell PRIOR to vSphere 5.1 as follows:
Both hosts are configured the same.
Network Adapters: vmnic4 and vmnic5
vSwitch
* MTU 9000
* Active Adapters vmnic4, vmnic5
* Load Balancing: Route based on the originating virtual port ID
VM iSCSI Network
* Active Adapters vmnic4, vmnic5
iSCSI2
* MTU 9000
* Active Adapter vmnic5
* Active Adapter vmnic4
iSCSI1
* MTU 9000
* Active Adapter vmnic4
* Active Adapter vmnic5
Storage Heartbeat
* MTU 9000
* Active Adapters vmnic4, vmnic5
With the enhancements provided in vSphere 5.1 is it safe to simply delete the SHB VMKernel port on each host?
swoods79
6 Posts
0
January 28th, 2013 08:00
My apologies, that was a typo. On iSCSI2 vmnic5 is active while vmnic4 is unused. On iSCSI1 vmnic4 is active while vmnic5 is unused.
Thank you for the quick response! Also thank you for the other information you've posted within the forum relating to setting LRO to zero, changing settings with respect to Round Robin, etc.