Question about HA device on 2 node Hyper-V Cluster

aciddo · Sat Jul 14, 2012 3:23 pm

Hello,

I am trying Starwind iSCSI SAN for Hyper-V on a 2 node Hyper-V Cluster.
I have latest version of starwind and free Hyper-V 2008 R2 with the latest updates.

I've created 2 HA devices (images) on Starwind (1 for Quorum and 1 for the VM) on both nodes.
I've connected to these devices using iscsi initiator (iscsicpl) on both nodes with MPIO option enabled.Also run mpiocpl and enabled MPIO on both nodes.

So far so good, both devices are recognized on hyper-v cluster manager where I have enabled CSV.I created then a new HA virtual machine using the cluster manager.
Now the problem is that when I try to simulate a failure on the cluster for example shutting down the first node the clustered ip does not appear on-line on the second node.Instead when I try to shutdown cluster service on each node at a time the clustered ip appears on-line on the second one.
I first thought that it was a problem on the cluster configuration and I recreated it but with the same results.
Trying to troubleshoot I noticed that when i cold shutdown one node the iscsi initiator on the second node does not mount correctly the disks on starwind SAN.So if i go and press manually "Auto Configure" the disks then magically the cluster starts to work on the second node! Isn't this supposed to happen automatically? I mean Auto Configuring the disks after loss of one node?

One second thing I would like to ask is what happens if for some reason one of nodes goes down and I have just a node now for the cluster and the HA devices on starwind management console appear with red exclamation mark (broken).Is there a way to make them "on-line" without bringing up the second node to do a full sync?

Thank you

Mon Jul 16, 2012 10:39 am

So far so good, both devices are recognized on hyper-v cluster manager where I have enabled CSV.I created then a new HA virtual machine using the cluster manager.
Now the problem is that when I try to simulate a failure on the cluster for example shutting down the first node the clustered ip does not appear on-line on the second node.Instead when I try to shutdown cluster service on each node at a time the clustered ip appears on-line on the second one.
I first thought that it was a problem on the cluster configuration and I recreated it but with the same results.
Trying to troubleshoot I noticed that when i cold shutdown one node the iscsi initiator on the second node does not mount correctly the disks on starwind SAN.So if i go and press manually "Auto Configure" the disks then magically the cluster starts to work on the second node! Isn't this supposed to happen automatically? I mean Auto Configuring the disks after loss of one node?

Would you please clarify where exactly are you pressing "Auto Configure" button? Maybe you mean Automount in iSCSI initiator?
Also, can you confirm that the CSV is online when you are shutting down one of the cluster nodes?

One second thing I would like to ask is what happens if for some reason one of nodes goes down and I have just a node now for the cluster and the HA devices on starwind management console appear with red exclamation mark (broken).Is there a way to make them "on-line" without bringing up the second node to do a full sync?

To be honest this sounds weird to me. I will explain: if you are shutting down one of the nodes, second one should stay inline and in SW Mgmt console it should appear with yellow exclemation mark, that means "Device is out of Sync", that is obvious, because second target of the HA is down. So, generaly speaking - if one SAN server failed - second one should serve as source for the data, that is gurantee 99.999% of uptime.

aciddo · Tue Jul 17, 2012 8:06 am

Hello again,

a) I mean when first node goes down, on the second node I have to start iscsi initiator (iscsicpl) and then go to "Volumes and Devices" and press the "Auto Configure" button.When this happens CSV successfully comes up.Yes I can confirm that CSV was on-line when I shut down the first node.

b) I understand, that is the normal behavior but when I was trying to simulate a node fail there was a condition that the second node was marked in red and I could not mount the volume.They only way was to bring the first node online again and do a full sync.Anyway I will to try to do more tests and report back.

Thank you for the assistance.

Thu Jul 19, 2012 10:29 am

Ok, I`m really confused now. Here are two of your quotations:

Yes I can confirm that CSV was on-line when I shut down the first node.

for some reason one of nodes goes down and I have just a node now for the cluster and the HA devices on starwind management console appear with red exclamation mark (broken).

One of those qotation can be truth only if second is false. I mean, or your HA device is broken, or the CSV, that is hosting on it is online.
Would you please clarify that please.

Thanks