Snapshots Take a Long Time When “Keep Memory” is Enabled

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms:

When taking a snapshot with memory, the VM may be unresponsive and the snapshot may take a long time to complete. This is because the ESX host has to dump the VM’s memory to disk.

When looking in the vmware.log file, we will notice that during the snapshot creation something called Lazy CheckPointing is utilized. Lazy CheckPointing is a feature that allows the VM to continue running while the memory is dumped.

Environment

VMware vSphere ESXi 6.5
VMware vSphere 6.7.x

Cause

By default mainMem.ioBlockPages is set to 64 pages. Each page is 4k, so 64 would be 256k and 2048 pages would be 8MB. The ioWait is set to 25ms by default. The array will preform differently with different write sizes, so you should test to see what works best for your environment.

Resolution

By increasing the ioBlockPages and decreasing the ioWait values, we can cut down on both the number of iterations and waiting time ( which in turn will decrease the total time to dump the memory to disk).

Add the below parameters based on your environment and load.

mainMem.ioWait
mainMem.ioBlockPages

It should look like below (value can change based on customer environment.

cat test.vmx | grep -i mainMem 
mainMem.ioBlockPages = "2048" 
mainMem.iowait = "0" 

Workaround:
There is no workaround, customer can use quiesce snapshot instead.