Running UCS as a ESXi VM

UCS - Univention Corporate Server
#1

Hello,

my UCS is running as a VM with ESXi 6.5. It becomes unproductive after a while. When i open the Overview in the VM Webinterface i can see that the RAM consumation is at 60%. A reboot bring it back to 40% but the system is very slow. The other VM on the ESXi are slow, too. I have to shutdown the VM and power on it again to bring it back to normal work.

I think it has to do something with the samba folder. In the night there is a backup script that copies all files to a NAS. When i start the script RAM consumation increases and don’t come back to normal level.

Do you have any idea how to finde the problem?

Kind regards

Jochen

#2

Hi,

it is false thinking the RAM usage should go down after it has been used very intensive. Additionally, usage is not an indicator of too few memory.

So let’s start from the beginning:

What do you mean by “becomes unproductive”? As your second VM is affected, too I would assume it is not an issue of UCS instead of ESXi.

So how much memory do you have physically in your ESXi host? How many VM do you have up-and-running with how much virtual memory assigned each?

/CV

#3

Hi,

thank you for your reply. Here are some answers:

  • The host has 16GB physical memory
  • There are 5 VMs
  • UCS has 12GB virtual memory (after Startup 3,3 GB in use)
  • 2 W7 PCs have 8GB each (6,5 GB in use)
  • 2 Linux VM have 3GB together (0,7 GB in use)

“Become unproductive” mean i can log in but can’t open “User” or “Software” for example. The startup takes about 5-10 minutes instead of 40 seconds when i power it on.

Kind regards

Jochen

#4

Hi,

so you have 16GB of physical memory?

And if I calculate correctly you have 12+2*8+3 assigned to your VMs? And they are all up and running? With 31GB of virtual memory? So nearly the double amount as physically available?

Well, in terms of vmware this is called overcommitment. Honestly, an overcommitment with factor 2 I have never ever seen working…

There are loads of articles regarding memory overcommitment in vmware (and hypervisors at all).
As abstract without educating too much here: If there is not enough physical memory available the memory “above” will be realized as swap space on the disks. Which is ways slower, of course. So around 50% of your VMs memory resides on the hard disk (and the VMs are not aware of it, how should they?). I guess you now know WHY your VMs are so slow.

So my recommendations here are:

  1. Reduce your VMs memory! If you want to have an acceptable performance do not overcommit or at least not about 100%. Acceptable values might be 20-40% but this depends on your VMs profile.
  2. Install vmware tools, if not done yet. Then the ESXi host may use a technique called “ballooning” to request VMs from within VMs.
  3. Tell the ESXi host to store the VMs swapfile on a VERY fast disk (if available), at least SSD, not on slow rotating HDDs.
  4. Buy more memory for you ESXi host!

No other conclusion here, sorry.

/CV

Mastodon