Hi, I have a couple of questions around memory reclaimation that I'm hoping someone can help with.
I think I understand the 4 memory states
high - no memory reclaimation
soft - ballooning as host approaches soft threshold
hard - as above, plus swap, compression (only happens on swapped pages?) ...(does break down of large memory pages into small , as so also TPS, happen here as well?)
low - as above, no further memory allocation to vms
I've seen hosts with high memory state, but with growing balloon and swap values (ie MCTLSZ growing towards higher MCTLTGT value and SWCUR growing towards higher SWTGT value), with no memory limits on the VMS. Anyone any ideas why this would happen?
I've also seen the following...
If I power on a VM on a host with no other VMs on it (VM has 3Gb memory, host has 146Gb), when the VM starts, it has 2Gb of shared memory.
To me this suggests that the host has had a shortage of memory in the past and has broken large memory pages into smaller ones, and run the TPS algorithm to create the hash table. When the VM is powered on, it's allocated small memory pages which are compared against and found in the hash table , therefore the pages are marked as shared ?
Does this mean that once a host has gotten to the stage where it's large memory pages are broken into smaller ones, they stay small until the host is restarted ? Is way to change them back to large memory pages ?
Thanks very much for your help,
Nick
