Skip to main content

Hello.

I have a question.Resource Requirements Supporting Snapshot Frequency (Asynchronous, NearSync, and Metro) I'm inquiring because I don't understand well even if I look at the link above.

Three G9 nodes
I'm thinking of making three G7 nodes into one cluster.
The G9 node already has a cluster configured and will expand the G7 node.
The G9 node is two SSDs and four HDD hybrid HCI.
G7 nodes are also six SSD All Flash nodes.

G9 Node Disk Capacity
SSD = 3.06Tib
HDD = 6.9qTib

G7 Node Disk Capacity
SSD = 1.92Tib

DR is configured in the G9 cluster that is already configured, and snapshots are being taken every two hours and sent to DR. The number of snapshots stored is two.

When configuring these services, how much CVM memory should be allocated?

The reason for putting the inquiry is that I set the CVM memory to 32GB with G9 initial configuration. It's been about 3 months since the configuration. CVM memory usage is over 90%.

We would like to ask for your opinion as the usage increases until a memory usage warning occurs in a short time.

Was the CVM memory allocation small in the first place?

 

Hi ​@cubensys 

Great question—and you're on the right track by monitoring CVM memory usage closely.

Based on what you’ve described:

  • 3 x G9 and 3 x G7 nodes

  • Mix of hybrid and all-flash

  • DR snapshots every 2 hours (with 2 snapshots retained)

  • CVM memory currently set at 32GB

  • CVM memory usage is over 90% after ~3 months

Always confirm with Nutanix Support before performing any work on production

Here’s my take:

32GB of CVM memory might be on the lower end, especially in environments with:

  • Mixed node types (hybrid + all-flash)

  • Frequent DR snapshots

  • Increasing storage footprint

CVM memory requirements grow as:

  • The number of vDisks, snapshots, and IOPS increases

  • Your cluster expands in capacity and usage

  • More services are enabled (e.g., replication, data reduction)

Suggested next steps:

  • Consider increasing CVM memory to 48GB or 64GB, depending on your hardware capacity and support limits.

  • Monitor CVM memory pressure using top or ncli—look for swap usage and memory overcommit.

  • Double-check if features like compression or deduplication are enabled—they consume more CVM memory.

  • Make sure each node’s CVM has the same memory to maintain cluster balance.

Bottom line:

Yes, 32GB is likely too low for your current config. I’d recommend scaling up before you hit a critical threshold.

Hope that helps—let me know how it goes!

 

Terms & Conditions