This seems to be normal with the MD-series SANs, or at least the MD3000-series. The failover time is over 2 minutes from what I've seen/experienced.

Dell's Equallogic range is more like 20-30 seconds. I'm not sure about Dell's Compellent series, but their Dell|EMC range with fiber channel had (Dell stopped selling/offering EMC) failover times within a few seconds or less.

Hi,

sorry for the formatting of the first post, my iPad did not correctely posted the message to the forum. Thank you for your reply. I have just another question, do you know if there is a possibility to decrease these timeout values?

Dell MD3000i, is this really an enterprise valuable storage for using with an hyper-v cluster?

Did you see somebody using Microsoft Multipath without Dell DMS file ==> Timeout less than 2 minutes?

Thanky you,

Gillato

-----------------------

Here the post correctely formatted:

Good Evening,

We have an Hyper-V Cluster composed of 2 nodes (Window
Server2008 R2 SP1 Enterprise - Hyper-V Role installed). Each node has 6 NIC's,
2 are used to do iSCSI with Dell MD3000i Storage. It is a dual controller
storage with 2 NIC's per controller. We used only the 1st controller to connect
our ISCSI MIC's from the cluster. We used 2 different subnets on Dell iSCSI
configuration.

 

Everything is working fine so far,traffic flows over both
NIC's from Hyper-V Cluster. We used Dell configuration Tools to configure each
host, so iSCSI is configured on each host with the following settings:

 - Multipath

 - ACTIVE/ACTIVE
paths with subsequent load policy

 

Problem: When we disconnect the first iSCSI NIC on node1,
it takes up to 3 minutes that the iSCSI initiator redirects all traffic over
the second iSCSI link on node 1. Both links stay ACTIVE/ACTIVE with the
disconnected link in reconnectin mode. During the 3 minutes, all VM's on node 1
are not operational.

 

I think the problem results from the fact that the
disconnected link stays ACTIVE => reconnecting. So the cluster sees 1
disconnected link and 1 online link => cluster status ok. But the multipath
driver does not take the disconnected link offline, at least not until 3
minutes are over.

 

Is this configuration supported for redundancy mode?

Do there exist some tuning settings to reduce these
timeouts?

What is the standard iSCSI configuration and how long
does failover generally takes using Dell MD3000i?

Should we use a different multipath policy (failover)?

 

I will thank you very much for each little hint to solve
this problem.

 

kind regards,

 

Gillato

Good Morning,

I have contacted Dell Support and they told me that you must connect all interfaces from the 2 controllers to have full Multipath Support. I have redone the configuration as follows:

 

Server interface 1 ==> MD3000i Controller 1 - interface 1
Server interface 1 ==> MD3000i Controller 2 - interface 1
Server interface 2 ==> MD3000i Controller 1 - interface 2
Server interface 2 ==> MD3000i Controller 2 - interface 2

But this does not help, timeouts stay at about 2-3 minutes when you disconnect an interface. The Dell Support told me that timeouts will only be 5-15 seconds when all interfaces are configured and connected.

Is this true now or will the timeout stay at 2-3 minutes? Is the timeout value limited by iSCSI protocol?

Best regards,

Gillato

Just specifically ask Dell support what the failover time is on your MD3000i (how long does it take for a virtual disk owned by controller 0 to fail over to controller 1 if you lose communication with controller 0).

This isn't dependent on the exact iSCSI setup, but to have the best chance to 'survive' this failover time, it's best to have a connection to all 4 ports on the SAN (though a minimum of 1 per controller is the actual minimum requirement).

Thank you for your answer. I am just not quite understanding exactly whate you mean with connections to four interfaces. The problem we face is that when our server has 2 network cards. The 2 network cards are on 2 different subnets.

So each network card has a connection to each controller.

Network card 1 (192.168.1.1) ==> Controller 1 Port 1 (192.168.1.10)

Network card 1 (192.168.1.1) ==> Controller 2 Port 1 (192.168.1.11)

Network card 2 (192.168.2.1) ==> Controller 1 Port 2 (192.168.2.10)

Network card 2 (192.168.2.1) ==> Controller 2 Port 2 (192.168.2.11)

When I now disconnect my first network interface on my server (192.168.1.1), the virtual disks failover on the MD3000i should not failover to the second controller because Server network interface 2 still has a connection to the 1 controller, but the timeout is still 3 minutes before the 2nd network controller on my server is sendind traffic over the interface.

Is this setup correct as I described it above or should I make the connections differently on my MD3000i.

Thank you very much in advance.

Gillato