Real failover problem with EVA8400

[jamesrico]

Hi,

I'm running SRM 4.1.2-484598 with two EVA8400's and 4.1 hosts.

During a 'real' failover test a couple of week ago the hosts running protected VMs became unresponsive during the failover of the first set of VMs. The remaining VMs on the hosts were unaffected from a network connectivity point of view. The hosts disconnected from vCenter and I could not connect a VI client directly to them.

The remaining failover plans were unable to gracefully shutdown protected VMs so when booted in the recovery site they were obviously in a crashed state; which was not ideal for a planned failover.

A reboot to the protected hosts brought them back online. The actual failover (and failback) completed sucessfully for 150+ VMs but its just wasn't very controlled!

Any idea? Sounds like a APD problem to me, I'm wondering if the LUNS were not correctly changed to 'Read Only' so to allow the protected hosts to still access inventory.

Thanks

[jamesrico]

Anyone?....

[stu161]

Hi,

I have exactly the same issue, with EVA 8400's

Testing failover to recovery site, causes all hosts on protected site to go into APD condition after rescan of datastores.

A restart of management agents on the hosts will rectify ... or a restart of host.

I currently have a case in with VMWare support for this... will post answer once fix  known  and tested.

Cheers

[stu161]

no fix for this problem..

VMWare will not publicly announce this, but this is a known bug, fix is to upgrade to V5.1

I have only seen this with the 8400's, no other storage, so this issue may be particular to this..

We are running this in production, and have had a site failure and failover/ failback, works well, apart from hosts going offline after failover(back)

Workaround is to reset management agents via SSH...  or DCUI.. but SSH works best... not a problem as long as you are prepared for it, and incorporate into your procedures.

Cheers

[jamesrico]

Thanks. We have upgraded to 5.1 and test failovers appear to be okay.