HA 5.0 - what if not all high priority VMs start?

[rickardnobel]

If using vSphere 5.0 with the new HA and I have selected for example 10 VMs with high priority and the rest with default medium, then after at failure the medium priority VMs should not start until all high-VMs have been restarted.

However, if for any reason a high priority VM fails to start - what will happen with the medium? Will they wait forever or is there any timeout value that will allow them to actually come up?

Also, if all high VMs do successfully restart, when are the medium VMs typically started? Is there any time value or similar that governs this?

[PduPreez]

Hi

There are a few changes in HA 5 regarding startups, but its on priorities and Retries.

Check out this post from Duncan Epping: http://www.yellow-bricks.com/2011/07/27/ha-architecture-series-restarting-vms-45/

To Answer your question: HA will try and start high priority VMs first, If there is any failure there will be retry attempts: Refer to link

However in the mean time all the other VMs will be started based on priority.

Remember that HA is not a guarantee, you could prioritize the VMs but cannot guarantee that the High priority VMs will be up before the Medium priority VMs, it also depend on the time for the OS to load for different VMs. Although HA will try its best.

So i short, just using the High, Med, Low priority. How I see it, and i do not have exact time delays, but my point of view.

Ha will issue start on all High Priority, 30sec later it will start all Med, 30sec later all low will start.

The VMs that fail to start will continue with the retries in their own time.

Note: This is exactly the same behavior as in vSphere 4

Please award points if you find this helpful/correct

[rickardnobel]

Thanks a lot for your response!

Pieter du Preez wrote:

Ha will issue start on all High Priority, 30sec later it will start all Med, 30sec later all low will start.

The VMs that fail to start will continue with the retries in their own time.

Note: This is exactly the same behavior as in vSphere 4

But in vSphere 4.x I do not think there was any delay between the startup of VMs of different categories (if there are available resources), but this was a new feature of vSphere 5.0, that you could prioritize different VMs and make them startup in a somewhat predictable order.

So it would be very interesting to know the amount of time between the startups and what will happen if one high priority fail to load at all.

[admin · Accepted solution]

Prior to 5.0, HA would restart all vms more or less at the same time. The restart priority was really only used to decide _which_ vms to failover if there are not enough resources to restart all of them. In 5.0, HA will restart all high priority vms before starting to restart lower priority vms. This ordering is only with respect to the "power on" operation. As soon as all the high priority vms are powered on, HA will move on to restarting the medium and then the lower priority vms. The high priority vms may still be booting up when the medium priority vms start getting powered on. Once HA has finished restarting high priority vms it immediately moves on to the lower priority ones - it doesn't wait for 30 seconds. And as mentioned already, this ordering is only for the initial failover attempt. If some high priority vm fails to failover for some reason, HA will not hold up failover of lower priority vms so the retry of restarting the high priority vm may occur while or after lower priority vms are being failed over.

Elisha

[shishir08]

VM restart  policy is always relative.High priority ones will be restarted first,and continuing to those with lower priority until all virtual machines are restarted.
Please note thatIf the number of hosts failures exceeds what admission control permits, the virtual machines with lower priority might not be restarted until more resources become available.

[PduPreez]

Hi Rickard

You are correct there was no delays in 4.x and neither is there in 5, Like I said I'm not sure if there are time delays, It's just a way to explain that the restarts happens (my point of view) in "POD's" and what ever fails to start will be started later and retry for 5 times. I might have confused the issue

Rickard wrote:

So it would be very interesting to know the amount of time between the startups and what will happen if one high priority fail to load at all.

See it as a big list of VMs sorted by priority from top to bottom.

HA will go down the list starting each VM. If 1 VM fails to start HA will carry on down the list and start VMs as long as there are enough resources available using your Clusters HA Admission control setting. There is no specific amount of time between startups.

If a high priority VM fail to start after 5 retries (not start at all) it will not affect HA process.

Elisha - also summed it up quite nicely

Hope this helps :smileycool:

[rickardnobel]

Thanks for the information!