iSCSI initiator connection failure

Hi Donald,

I already have a support case open (regarding VMs freezes), I collected many and many logs and your colleagues told me to:

1) enable jumbo packets on physical network cards

2) disable virtual machine queues on iscsi virtual network cards

unfortunately, the messages still persist.

SAN and hosts are connected to 2 dell SFP+ stacked-switches 10gig, and we use DCB too.

Can you please confirm that getting that message is something NOT expected ?

Is it worth to update also broadcom network drivers ?

Thanks for help.

Hello,

That error is know as a KATO (Keep Alive Time Out).   Periodically, the array sends a message (Keep Alive) to the initiator.   It functions very similar to a simple PING verifying conenctivity.   If that fails, the array will drop the connection.  This is per the iSCSI spec.  The initiator should detect the condition and log back in.   The initiator too does this same thing, sends the arrays a "ping" as well.

Many times this points to issues at the network layer.  

I would suggest opening a support case.  They'll need array diags, switch logs and SANHQ archive if you have that installed.  If you don't then you should consider it.

If you haven't already upgraded to 7.0.9, it would be a good idea to do so.

Regards,

DELL N4032F switches.

I make connections using MS iScsi initiator but I don't get very well the difference between CNA and ms initiator, also because the setup was lead by some consultants here.

Thanks for help.

Yes, that is NOT a normal message.  

What model switches?    

I don't like to make suggestions when you already have a case open.   It can just add confusion to the process.  Since I don't have the logs, etc...

Are you trying to directly connect to the EQL array with the MS iSCSI initiator, or are all the conenctions being handled by the Converged Network Adapter (CNA).   Sounds like your using the Broadcom adapters.

The only version of Windows that I am aware that claims any kind of support for DCB with their SW iSCSI initiator is 2012R2.   I've not actually seen it working however, and has not been qualified with EQL.  Which means any iSCSI traffic using that will be in the "Lossy" class of service.

Are these switches dedicated to iSCSI use?    If so, then I might try not using DCB and just use it as a standard.  With a dedicated switch you don't need to use DCB.

The SW iSCSI initiator in Windows 2012R2 has not been qualified for direct use of DCB.  Nothing to do with the switches.   For DCB support I have only seen a Converged Network Adapter (CNA) which fully offloads iSCSI, DCB and networking from the host work.   Windows has added DCB support within their network and iSCSI stack, but I have not seen it work in practice.