How do I know if I've reached my capacity of number of VMs?

[carwyn]

We have 2 ESX servers and a AX150i SAN. There are currently 26 VMs in use.

Looks like the servers are loosing connection with the SAN, one of our developers tried copying a 2GB file off one of the VMs earlier and this resulted in all VMs on this ESX crashing, with no heart beat. I tried to reset the VMs, but this didn't work so had to reboot the ESX host.

I am starting to think that we have reached our maximum capacity. But does anybody know how to check this or have any real world advice on AX150i in terms of number of VMs.

Thanks

[alanrenouf]

I may be wrong but it sounds more like a potential issue with the way you have your networking setup, Are all your vm's on one vSwitch ? how many nics do you have in your VM Virtual switch, how are these configured and what speed are these ?

If you found this information useful, please consider awarding points for Correct or Helpful.

Alan Renouf

VMware, Citrix, Microsoft Consultant

UK

[carwyn]

See attached jpeg there are 3 vSwitch and 4 Nics on each Host.

I used two nics for the iSCSI connection for improved performance and redundancy.

iSCSI network is also in a seprate VLAN.

It's been ok like this for 12 months now?

Is this setup incorrect?

[alanrenouf]

Ideally I would seperate the service console from the VM network and have dual 1000MB cards on both.

Also, do you realise you have 18 VMs sharing a 100MB connection on your WAN vswitch, is this by design ?

Do you have a spare slot in the server you can put an extra 2 or 4 port nic in ?

[carwyn]

Thanks, yes the WAN is set at 100mb by design.

Isn't the service console and the VM network already seprate?

I could add another 1000mb card and add it to the LAN network, I'm not convinced it's a LAN bottleneck problem.

Why would this freeze all the VMs?

[alanrenouf]

Your image shows LAN and Service console on the same Nic

[carwyn]

Yes so it is, sorry. I will go ahead and buy 4 port Nic for both ESX hosts and give this a try.

How sure can you be this is the cause. I don't understand why speprating them helps, do they need to be on different vlans as well?

[TomHowarth]

there is more to it than pure networking, but that is for another conversation, but consider this, in the physical world those 26 Guests that you have virtualised on that switch would each be sitting on a single or greater amount of Network cards.

personally I would as suggested by Alan add another 4 port "supported NIC card to your host/s and firstly seperate the Service console from the production network. then I would retire the 100Mb NIC from your DMZ network

You will now have 7 NICs to work with

I would configure then as follows

NIC 0 SC fail over to VMkernel

NIC 1 VMKernel failover to SC

NIC 2 + 3 DMZ Network

NIC 4-7 Production network

This will give you resillience and also a better security if you are using VLANS in your environment.

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

Tom Howarth

VMware Communities User Moderator

Blog: www.planetvm.net

[carwyn]

I will recomend the extra ports to my boss. I am still unsure as to why this will stop the host from crashing. Would a LAN bottleneck cause a ESX host to crash?

I've added some esxtop screenshots, one shows normal operation the other whilst I was was download a 1GB file. No idea if this helps, but I would be greatfull if you experts could take a look and see if anything leaps out to you.

Thanks

[espi3030]

Definitely sounds like you have a setup issue and have not reached a limit. For configuration maximums see this -->

[carwyn]

Ok so I can use 64GB of memory & 32 nics on the host if I wanted. Intresting.

But why would you say it's setup issue?

Is my setup wrong?

Do you think like the other that I need to add a 4 port Nic and seprate the service console from the LAN?

Thanks

[carwyn]

It also says I can have a maxmium of 200 VMs per host. So I guess in term of my original question, that document was usefull.

Thanks, I will give you one browny point.

Just need to know why my configuration is possibly wrong now.

[TomHowarth]

200 VM's per host, I always find the an amusing section considering that you can only have a maximum of 192 vCPU's per host. so we have 8 VM's with no processors LOL

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

Tom Howarth

VMware Communities User Moderator

Blog: www.planetvm.net

[espi3030]

Well from reading your inital post, doesn't necessarily sound like your setup is wrong per se.

How are you copying files to from VM's? From both esx hosts try a vmkping to the vmotion IP to each other, see if you get good returns.

[carwyn]

Copied from the VM client to my laptop via the UNC path like - copy to c:\

The stange thing was I could ping ok, althought the response time were a bit high. The VMs were still ON and the heart beats had stopped.

This "copying a large file and then the ESX host crashed" thing however could be a red herring. I'm going to check the network activity again on previous crashes to see if the same happened.