Unsolved
This post is more than 5 years old
2 Posts
0
9952
August 17th, 2011 03:00
vSphere Network Failover on M710HD blades with two 10GBe NICs
Hi All,
I would appreciate your input on best practices for network redundancy in our scenario.
Our setup is as follow:
M710HD Blade servers with 2 x 10GBe NICs, each NIC connected to separate blade mounted switch (M8024k). The switches are not stacked and operate as separate devices with 2 x 10GBe link between them. Then there is an uplink from each switch to the distribution layer on the network (of course one link is disabled with STP to avoid a loop).
vSphere 4.1 ESXi. We're using vDS with number of PortGroups for VM traffic, Management VMotion, iSCSI etc. andNetwork I/O Control Enabled.
What I've noticed is that in this scenario, standard vSphere Network Failover Detection of "Link Status Only" doesn't work. The link is presented as UP until a switch is physically pulled from the chasssis. So this does not cover majority of failures on the switch or even reboots. If switch stops responding vSphere would not initiate failover to the other NIC, until the faulty switch is physically pulled out from the blade chassis.
I understand that vSphere Beacon Probing is not reliable with two physical NICs, so what is the answer here?
Thanks
Wojtek
I would appreciate your input on best practices for network redundancy in our scenario.
Our setup is as follow:
M710HD Blade servers with 2 x 10GBe NICs, each NIC connected to separate blade mounted switch (M8024k). The switches are not stacked and operate as separate devices with 2 x 10GBe link between them. Then there is an uplink from each switch to the distribution layer on the network (of course one link is disabled with STP to avoid a loop).
vSphere 4.1 ESXi. We're using vDS with number of PortGroups for VM traffic, Management VMotion, iSCSI etc. andNetwork I/O Control Enabled.
What I've noticed is that in this scenario, standard vSphere Network Failover Detection of "Link Status Only" doesn't work. The link is presented as UP until a switch is physically pulled from the chasssis. So this does not cover majority of failures on the switch or even reboots. If switch stops responding vSphere would not initiate failover to the other NIC, until the faulty switch is physically pulled out from the blade chassis.
I understand that vSphere Beacon Probing is not reliable with two physical NICs, so what is the answer here?
Thanks
Wojtek
No Events found!
mikemooney
11 Posts
0
August 18th, 2011 10:00
One thing that can be done with the M8024s is to configure them in 'simple mode', you would then configure your 10GB uplink as a 'Port Aggregator Group', you can then aggregate multiple uplink into a single group and set a 'minimum uplink' on the group, if a link fails the group will detect this and initiate a fail over or a shut-down procedure.....
The M8024 documentation details how this can be done.......
Mike.
ANOC
3 Posts
0
December 1st, 2011 11:00
Hi there,
I have exactly the same issue. I managed to get it running using beacon probing as you describe, and it does seem to be working ok sofar.
My only suggestion is apparently 4.2 firmware is due in a couple of days (7th) which may look to help with this problem.
For what it's worth, dell really doesn't seem to understand why this is a problem.