FT won't enable for vm

[wb2]

I just tried turning on Fault Tolerance on one of my virtual machines.  It's a single processor, 6gb ram.  We're running ESXi 4.1 with 2 hosts.

I get this error:

Insufficient resources to satisfy configured failover level for HA.

Anyone know why?  It's the first machine I've attempted to set FT on.

[Troy_Clavell]

With 2 Hosts, you may want to set you admission control level to % of... instead of Host failures.  How is your Admission control policy setup for HA?

[wb2]

Host failures cluster tolerates: 1

[Troy_Clavell]

try using % instead, and see what happens.

[wb2]

What exactly is the difference?  I've never used the % before.  What do you recommend for a number.  It's set to 25% right now but greyed out.

[Troy_Clavell]

Percentage based admission control uses a more flexible mechanism.  It accumulates all reservations and subtracts it from the total amount  of available resources while making sure the specified spare capacity is  always available.

http://www.yellow-bricks.com/vmware-high-availability-deepdiv/

However, if it's greyed out, you may have admission control disabled, is that the case?  I would use the default 25%

[jjkrueger]

Changing your HA cluster settings to a Percentage Reserved will likely resolve the problem, as Troy said.  With a 2-host cluster, I'd set the reservation to 50%, just to ensure failover capacity.

What you're running into is a combination of the nature of HA capacity planning when using the Number of Host Failures Tolerated option and the way FT works.

When you enable FT on a VM, a reservation for 100% of the memory is placed on both the primary (and the target secondary) VMs.  In your case, 6GB.

When you specify a number of host failures to tolerate for HA, it will look at your running VMs and determine a slot size.  You can view this slot size by selecting your cluster in your vCenter inventory, looking at the HA box in the Summary tab, and clicking the Advanced Runtime Info link.

HA calculates this slot size by looking at the memory and CPU reservations on each running VM in the cluster, and taking the largest values.  In your case, with FT enabled, your slot size will be set for at least 6GB of RAM (a little more for VM memory overhead).  HA then makes a decision that, in a worst case scenario, _every_ VM in your cluster will need 6GB of memory to run.  For example, if you have 10 VMs, HA will plan its worst case to require 60GB of memory to accommodate the failover load.  That would mean 60GB per host in a 2 node cluster.

Changing the HA failover setting to a percentage of resources means that HA claims a reservation on a percentage of the total CPU and memory resources in your cluster to deal with failover capacity, so the slot size calculation is no longer necessary. 

The writeup at Duncan's blog (that Troy linked) is a fantastic read on the subject.

Hope that helps,

-jk

[bulletprooffool]

FT has a long list of requirements and limitations - things like limit in number of vCPUs, dedicated Gigabit networks etc.

Have a look at :

http://communities.vmware.com/blogs/vmroyale/2009/05/18/vmware-fault-tolerance-requirements-and-limi...

good luck and pleas efee dback if you solve your problem

[wb2]

Thanks for the great replies.  I set to 50%, then enabled FT again on the VM.  I get this in a "Fault Details" window.  The unused disk blocks of the virtual machine's disks have not been scrubbed on the file system.  This is needed to support features like Fault Tolerance.

Anyone know what this means?

[Troy_Clavell · Accepted solution]

see if this helps

http://kb.vmware.com/kb/1021615

[wb2]

I was actually just reading that article over.  I've never heard of lazy zeroed disks.  Does anyone know what this is?

[Troy_Clavell]

to use FT, you must have thick provisioned VM's.  You cannot use FT with thin provisioned VM's

[wb2]

The VM currently has "Hard disk1" and under Disk Provisioning it says:

Type: Thick

Provisioned Size: 60GB

Or is it referring to something else?

[Troy_Clavell]

see below snip from the Availability Guide:

You have attempted to turn on FT on a powered-on virtual machine which has thick formatted disks with the property of being lazy-zeroed. FT cannot be enabled on such a virtual machine while it is powered on. Power off the virtual machine, then turn on FT and power the virtual machine back on. This changes the disk format of the virtual machine when it is powered back on. Turning on FT could take some time to complete if the virtual disk is large.

[wb2]

Do you know what lazy zeroed disks means?  I'm just curious why it would refer to it in the error if my disks are thick provisioned?

[Troy_Clavell]

see below for all disk types, zero lazy is defualt

http://kb.vmware.com/kb/1022242

[wb2]

So I assume it'll convert it to eagerzeroedthick?

[Troy_Clavell]

wb2 wrote:

So I assume it'll convert it to eagerzeroedthick?

That is correct.

[jjkrueger]

There are two different kinds of Thick disks: eager zeroed thick, and lazy zeroed thick.

Lazy zeroed thick virtual disks are the default. When these are provisioned, the size of the disk is pre-provisioned, but the actual blocks in the disk are not zeroed out until the guest attempts to write to that block.

Eager zeroed thick virtual disks are created when you select the "support cluster features like FT" during the virtual disk creation. These virtual disks are completely zeroed out during provisioning, typically taking quite a bit longer to create.

Both will report as "Thick" disks in the vSphere client.

[wb2]

This is going to be tricky.  It's my vsphere VM that I'm trying to turn FT on.  How do I enable FT while the VM is turned off if I need vSphere to do this?