ESX hosts and bladecenters

[fgl]

Excuse me but my brain is drawing a blank. I remember something way back when bladecenters and esx hosts were coming together that if you had multiple bladecenters and more than 4 or 5 esx host in the same ESX cluster that you need to make sure that no more than either 4 or 5 of those esx host were not on the same bladecenter. It had something to do with being able to failover and that either HA or vmotion would fail because all the master something would have been on the esx hosts on the bladecenter that just died. I'm sorry if I can't explain it better but does anyone know what I am referring to and how to configure these so-call master host to be spread across multiple bladecenters.

Thanks.

[FredPeterson]

The first four hosts that join a HA cluster are considered masters of the cluster so those first four should be kept separate from each other as much as possible to ensure the cluster maintains operational status.

There is nothing particular you have to configure, just know that its the first four that join the cluster.

[chadwickking · Accepted solution]

Gotta love that when it happens??

Happens to me all the time

He is right, HA designates primaries and secondaries and that changes based on what affects HA, whether a host fails or if a HA is reinstalled. Duncan does a very good job of explaining HA and going into primaries and secondaries.

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

I also believe this is where I heard mention of the very same thing you are talking about. We do the same thing in our Datacenter - spread out the blades between chasis. Duncan I believe even goes further in telling you how to manually set primaries and secondaries.

Cheers,

Chad King

VCP-410 | Server+

Twitter: http://twitter.com/cwjking

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[unsichtbare]

I run multiple ESX on blades all the time, up to 16 ESX servers in an enclosure. The likelihood of an entire enclosure failing is significantly less than an individual rackmount server, due to redundant (6X) power supplies, cooling (10X) fans, interconnect bays (8X) for managed switches, etc.

Obviously, if one had 2 enclosures and was creating HA clusters, it would be advisable to configure ESX Servers on each enclosure to further reduce the possibility of hardware failure.

-J

[FredPeterson]

If all you have is one chassis the point about cluster primary/secondary status is moot.

The point was the first 5 hosts in a cluster are primaries. Having them all in the same blade chassis is a monumental mistake to the integrity of your cluster.

If all primary hosts fail simultaneously no HA initiated restart of the VMs will take place. HA needs at least one primary host to restart VMs. This is why you can only take four host failures in account when configuring the “host failures” HA admission control policy. (Remember 5 primaries…)

[fgl]

Thanks everyone especially chadwickking as that was exactly what I was looking for. I remember reading this a few years ago when we only had one bladecenter and I wasn't concern but over time and now we have 4 bladecenter with 48 blades I just needed to make sure we didn't have all of our eggs in one basket sort of speak. Thanks again.

[chadwickking]

No problem it is a pleasure to assist and be of use. Happy Virtualizing and have a good night!

Cheers,

Chad King

VCP-410 | Server+

Twitter: http://twitter.com/cwjking

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[rickardnobel]

>now we have 4 bladecenter with 48 blades I just needed to make sure we didn't have all of our eggs in one basket sort of speak.

As has been pointed out there are five primaries and at least one of them must survive for HA to work.

If you would like to really know, then enter the Service Console on one host running HA and go to the following directory:

/var/log/vmware/aam

And read the file "aam_config_util_listnodes.log", perhaps with:

cat aam_config_util_listnodes.log | more

You will be able to see which five hosts are primaries for this HA cluster. To be safe you should see that they are spread over at least two bladecenters. IF (not likely) all are on the same you could try to put one of the primary HA servers into maintaince mode. This should cause HA to select a new random one.

[Bigi201110141]

Are there any negatives of spreading HA Cluster Hosts between multiple BladeCenters as the Networking will be going out of the BladeCenter internal switch to another?

[TedH256]

only drawback would be if your chassis uplinks are saturated or otherwise drastically less responsive (ie higher latency, etc ...)