MPIO Network failure stops all traffic.

[ggalley]

In windows if I disable even a single network adapter on any server all traffic halts for 17 to 25 seconds is this normal is there a way to reduce this time or to not have it happen at all?

Configuration/Usage Below:
Starwind version 5.7.1807
Servers are all running 2008 R2 Enterprise
I have starwind configured in HA mode with 2x10GBE cabled to each switch.
The HyperV hosts currently have 4x1GBE dedicated to ISCSI.
This is on a dedicated storage network.

I have applied your recommended network settings to the starwind servers
http://www.starwindsoftware.com/forums/ ... t2293.html
I have jumbo frames enabled.
MPIO policy is set to use round robin.

Using Barts test stuff I run a stream of data to the mounted ISCSI drive.
All network adapters light up and I see network utilization on all storage server adapters and hyperv adapters.

Starwind 1
10.254.1.50 -> switch A
10.254.2.50 -> switch B
Starwind 2
10.254.3.50 -> switch A
10.254.4.50 -> switch B

HyperV 1
10.254.1.51 -> switch A
10.254.2.51 -> switch B
10.254.3.51 -> switch A
10.254.4.51 -> switch B

Are any of you familiar with this microsoft kb
http://support.microsoft.com/kb/2522766

[Anatoly (staff)]

Could you please provide us with detailed network diagram and specify what exactly NICs on what exactly servers have you disabled?

Thank you

[ggalley]

I have attached a picture of the network layout.

After I disable/unplug the single nic from store02. Every ISCSI network connection hangs for 10-30 seconds.

Is this to be expected with MPIO a single connection lost makes all clients stop sending traffice?

[Anatoly (staff)]

Dear ggalley,

Since you are using two servers I am assuming that you are using HA, so please see attached diagram with recommended settings (I haven`t specified all the IPs but I think you will get the idea):

[ggalley]

Is this Gap normal with MPIO ISCSI failovers after a single NIC failure?

NetworkIOGap.png (2 KiB) Viewed 8571 times

I made the changes you request I am still getting the same results. I have configured my ISCSI to enable multi pathing and have multiple sessions. I have also configured my iscsi without enabling multi pathing with the same results.

Network activity while I have Barts test stuff running from hyper01 connecting to a single ISCSI mounted drive.

NetworkIO.PNG (50.64 KiB) Viewed 8573 times

This is getting more urgent than a once a day response.

[Anatoly (staff)]

Dear ggalley,

First of all as I can use your ASM is valid so you are welcomed to use it (you can email us using support@starwindsoftware.com]this email), and you can expect that you will get response faster.

I would also like to ask you to clarify on what NIC exactly do you see the gap that you have specified?

[anton (staff)]

Run remote session with them and publish results here. Should save quite a time on turn around. Thanks!

Dear ggalley,

First of all as I can use your ASM is valid so you are welcomed to use it (you can email us using support@starwindsoftware.com]this email), and you can expect that you will get response faster.

I would also like to ask you to clarify on what NIC exactly do you see the gap that you have specified?