Memory impact on rolling upgrade | Nutanix Community
Skip to main content
Solved

Memory impact on rolling upgrade


Wing Chun

Hi all…

I have 3 node DX cluster with around 50 VMs powered on residing on AHV.

We are planned for a rolling upgrade, CPU and storage resources are good enough.

However, memory utilization is a concern:

Memory of powered on VM = 690GB

Memory of powered off VM = 210GB

Memory usage 91%

Total pRAM = 754GB RAM

Will this be risky for a rolling upgrade given that no pRAM is added and no VM planned to power off?

Best answer by JeroenTielen

Do you have HA reservation turned on?

I don’t think so, so there will be vm’s shutdown as there will not be any ram available when a node is rebooting. If you want to control it shutdown VM’s that can safely been shutdown leaving the critical running. 
 

Dont forget to turn on HA reservation as this will limit (or even warn you) when going over the memory limit for one node down. 

View original
Did this topic help you find an answer to your question?

6 replies

JeroenTielen
Forum|alt.badge.img+8
  • Vanguard
  • 1413 replies
  • Answer
  • April 10, 2025

Do you have HA reservation turned on?

I don’t think so, so there will be vm’s shutdown as there will not be any ram available when a node is rebooting. If you want to control it shutdown VM’s that can safely been shutdown leaving the critical running. 
 

Dont forget to turn on HA reservation as this will limit (or even warn you) when going over the memory limit for one node down. 


Hi ​@JeroenTielen  - When we have HA reservation turned on, does it prevent you from creating VM when it's going over the memory limit for one node down or where does it warn you.

 

Thanks,

Mahendra.


JeroenTielen
Forum|alt.badge.img+8

Sorry for the late reply. Yes when HA reservation is turned on you cannot start a VM when you go over the limit. It will create an alert. 

 

If you run LCM to update software, which requires a reboot, it will also show a banner saying there are not enough resources so virtual machines will be shut down. 


Thank you ​@JeroenTielen . Appreciate your clarification on this. Is there any thumb rule on creating a threshold for Memory utilization for production cluster. Example, warning alert at 60% usage and so on.


JeroenTielen
Forum|alt.badge.img+8

You can create custom alter policies in Prism Central. So that are warned at your own preferred threshold. 

 

Like this: 

 

 


Thank you so much!


Reply