Control Behaviour when Datastore Goes Offline

[J1mbo]

Hi,

If a VM has VMDK's provided on two datastores, and one of those datastores goes off-line, ESXi doesn't notify the guest and neither does it suspend the guest.

I wondered whether this can be controlled, for example to suspend the guest in this scenario?

Many thanks!

[Troy_Clavell]

you may be able to get some help with alarms.  Depending on the trigger, you can set an action to Power Off or Suspend a guest.

[Shakaal]

Hi,

Following is the default Guest behaviour in case of underlying disk becomes inaccessible, this is irrespective of whether the machine is a physical machine or virtual machine

For the following actions to happen the guest should issue a block/file level changes to the files system which requires disk access. if there is no disk activity the guest will do nothing.

Windows Guest OS will BSOD if the disk is inaccessible for 60Seconds

Linux Guest OS will make the filesystem Readonly, if the disk is inaccessible for 90Seconds

Procedure to increase the timeout value for windows

To increase the disk timeout value:

1. In the registry, go to HKEY_LOCAL_MACHINE/System/CurrentControlSet/Services/Disk.
   
2. Right-click and select Edit/Add value.
   
3. Set the value name to TimeOutValue.
   
4. Set the data type to REG_DWORD.
   
5. Set the data to 0x03c hex (or 60 decimal).
   
6. Reboot the virtual machine.
   

Please refer to below KB article for linux guest

http://kb.vmware.com/kb/1009465

Regards

[Shakaal]

also found following blog

http://www.boche.net/blog/index.php/2009/10/29/vmware-esx-guest-os-io-timeout-settings-for-netapp-st...

Regards

[J1mbo]

Many thanks for the replies on this.

I was considering provisioning a single drive to a VM that would itself then provide an NFS share to the main cluster (ISO/dev/sand-box space making use of a single free drive bay), but it seems in this case it's going to introduce too much risk to the rest of the system.  Will revert to a simple dedicated machine for the job.

As an aside, when I tested dropping an NFS share providing storage to an Ubuntu 10.10 VM, the guest didn't report an error until 1,080 seconds (even through the entries in 99-vmware-scsi-udev.rules are as indicated by the KB article).

[Shakaal]

I would suggest you to try VSA(vSphere storage appliance in 2node Cluster) for further information on VSA check the following link

http://www.vmware.com/products/datacenter-virtualization/vsphere/vsphere-storage-appliance/overview....

Regards