HP 4300 SAN ALB failover issue

[abovebrd]

HP 4300 SAN ALB failover issue

I believe ALB is designed and supposed to work differently than what I am currently experiencing in my testing.

We have a P4300 SAN connected to two Cisco 2960 switches. Each switch is connected to the network core through a trunk connection. Each NIC on each SAN is connected to one Cisco switch. So there are two SAN nodes and two switches. All VLAN operations appear to be working

We when power down one switch both SAN nodes remain connected, however when we unplug the switch trunk port one of the SAN nodes becomes unreachable. Now the interesting thing is when we disconnect the other trunk port of the other switch the other SAN node becomes unreachable while the other remains active.

Again , if one of the switches or Ethernet ports gets turned off the failover seems to work fine.

Is this behavior expected ? or could this be a design or configuration issue.

 

[abovebrd]

It looks like it was a design / config issue

It turned out to be a spanning tree problem. We had portfast enabled on the iSCSi vlan interfaces. Disabling portfast inconjuction with a trunk cross connect between both switches, resolved the issue.

Now when we unplug the trunk connection to the core on either switch both storage nodes stay up.

 

[kmcferrin]

This is a very common issue with networking failover (for iSCSI SANs, network teams, etc). The failover usually does not detect a failure beyond it's link to the switch. This means you can power down one switch, or unplug the link from the device to the switch and it will correctly fail over. But if you disconnect the upstream connection the failover software will not detect the link state failure.

________________________________________
CompTIA A+, Network+, Server+, Security+
MCTS:Windows 7
MCSE:Security 2003
MCITP:Server Administrator
MCITP:Enterprise Administrator
MCITP:Virtualization Administrator 2008 R2
Certified Quest vWorkspace Administrator

 

[abovebrd]

Our experiecne was really bizare (or so we thought) Unplug trunk to switch1 and SAN node 2 became unreachable. Then reverse the process, Unplug trunk to switch 2 and San node 1 became unreachable. However a power cycle or unplugged ethernet interface produced the expected failover result.

At first we thoguht we must have adaptive load balacing misconfiguring. But after several hours with HP support
we realized it was network issue.

We overlooked the cross connection because we initially thought it was a bit over engineering and why introduce potential switching loops (LOL) Leason learned

kmcferrin thanks for confirming