Collecting memory utilization metrics from KVM virtual machines

You can use the Agent to collect memory utilization metrics from KVM virtual machines. To enable the KVM hypervisor to provide these metrics, you must run the dommemstat command as follows:

1. Log in to your KVM host.
2. Verify whether the dommemstat version supports the --period <number> option by running the following command. This option is required for enabling the metrics.
   virsh help dommemstat

3. Verify whether the following results are displayed that indicate support for the --period <number> option:
   NAME

   dommemstat - get memory statistics for a domain

   SYNOPSIS
   dommemstat <domain> [--period <number>] [--config] [--live] [--current]

   DESCRIPTION
   Get memory statistics for a running domain.

   OPTIONS
   [--domain] <string> domain name, id or uuid
   --period <number> period in seconds to set collection
   --config affect next boot
   --live affect running domain
   --current affect current domain

4. Run the following command to enable the metrics:
   virsh dommemstat <ID> --period 10 --live
   where, ID indicates the ID or name of the KVM.

5. Verify that the metrics are enabled by running the following command:
   virsh -c qemu:///system --readonly dommemstat <ID| Name of VM>
   where ID is a domain name. To view a list of available domains, run the virsh list command.
   If the KVM is collecting memory statistics, it displays metrics similar to the following:
   swap_in

   swap_out

   major_fault

   minor_fault

   unused

   available

   usable

    If the KVM is not collecting memory statistics, it displays metrics similar to the following:

    actual

   last_update

   rss

6. Verify that both the available and unused metrics are displayed and show non-zero values.

   Additional examples

   To list current VMs:

   $ virsh -c qemu:///system --readonly list

   To check the memory statistics for all current running VMs:

   $ for ID in `virsh -c qemu:///system --readonly list | grep running | cut -d ' ' -f 2`;do echo "ID = $ID";virsh -c qemu:///system --readonly dommemstat $ID;done

   ID = 14

   actual 16777216

   swap_in 0

   swap_out 0

   major_fault 45249

   minor_fault 789565533

   unused 15374012

   available 16257500

   usable 15629308

   last_update 1622572158

   rss 1574076

    

   ID = 15

   actual 8388608

   last_update 0

   rss 8457856

    

   ID = 17

   actual 16777216

   swap_in 0

   swap_out 0

   major_fault 42413

   minor_fault 790621492

   unused 15348592

   available 16257500

   usable 15646232

   last_update 1622572157

   rss 1692932

   In this example, VMs 14 and 17 are collecting memory statistics, VM 15 is not collecting memory statistics (it is running on Windows). Note that VMs running Windows will not be able to be successfully configured to collect memory statistics.

   To configure any individual VM to collect memory statistics requires root permissions:

   $ su root -c “virsh -c qemu:///system dommemstat <ID|Name of VM> --period 10 --live"

   To configure all running VMs to collect memory statistics:

   $ su root

   $ for ID in `virsh -c qemu:///system --readonly list | grep running | cut -d ' ' -f 2`;do echo "ID = $ID";virsh -c qemu:///system dommemstat $ID --period 10 --live;done