Skip to main content

kvm ksm tuning

Context

I am running a no-brand machine, 

  • CentOS 6.3
  • 2 cores
  • 8GB RAM
  • 1 SATA disk

On this I fired 6 virtual machines

  • 2 Centos 6, 1GB RAM each
  • 4 Windows XP, 2GB RAM each

The reader probably noticed I overbooked RAM: If all the VM will use the available RAM it will fill up to 10GB.

I wanted to test the KSM effect on this. The default KSM settings on this OS is:

 KSM_MONITOR_INTERVAL=60
KSM_SLEEP_MSEC=100
KSM_NPAGES_MAX=1500

The result I got is:

  • Memory usage fits in RAM (very low swap usage
  • Too high CPU usage for scanning "same pages" (constantly using the equivalent of 70% of 1 core)
  • Disk IO is reasonnable, but on the other hand load average is high (10)
  • The Windows XP VM are desperately slow (not a surprise)

Tuning

Before tuning, I collected what ressource I could use more

  • This machine can bear more disk IO: I can use swap and RAM a bit more
  • This weak CPU should be a bit more used by VMs instead of KSM

My fisrt idea is to make KSM check less often. This is my new KSM setting:

KSM_MONITOR_INTERVAL=600
KSM_SLEEP_MSEC=1000
KSM_NPAGES_MAX=5000

The result is:

  • Less load average (now 6)
  • Less CPU used (now about 40% of 1 core)
  • More swap used, but IO statistics are still OK (IO Wait, IO throughput)

I'll still wait for users feedback before sticking to this setting (or not).

Comments

Popular posts from this blog

npm run build base-href

Using NPM to specify base-href When building an Angular application, people usually use "ng" and pass arguments to that invocation. Typically, when wanting to hard code "base-href" in "index.html", one will issue: ng build --base-href='https://ngx.rktmb.org/foo' I used to build my angular apps through Bamboo or Jenkins and they have a "npm" plugin. I got the habit to build the application with "npm run build" before deploying it. But the development team once asked me to set the "--base-href='https://ngx.rktmb.org/foo'" parameter. npm run build --base-href='https://ngx.rktmb.org/foo did not set the base href in indext.html After looking for a while, I found https://github.com/angular/angular-cli/issues/13560 where it says: You need to use −− to pass arguments to npm scripts. This did the job! The command to issue is then: npm run build -- --base-href='https://ngx.rktmb.org/foo&

dockerfile multiline to file

Outputing a multiline string from Dockerfile I motsly use a Dockerfile by sourcing from a base ditribution: CentOS or Debian. But I also have a local mirror and would like to use it for packages installation. Espacially on CentOS it is about many lines to write to the /etc/yum.repos.d/CentOS-Base.repo file. Easiest way: one RUN per line The first method that comes in mind is to issue one RUN per line to write. Here you are: RUN echo "[base] " > /etc/yum.repos.d/CentOS-Base.repo RUN echo "name=CentOS-$releasever - Base " >> /etc/yum.repos.d/CentOS-Base.repo RUN echo "baseurl=ftp://packages-infra.mg.rktmb.org/pub/centos/7/base-reposync-7 " >> /etc/yum.repos.d/CentOS-Base.repo RUN echo "gpgcheck=0 &quo

Gnome VNC Grey Checkboxes

When setting up VNC server in Gnome, I get a grey screen (or a black one) and the checkboxes to "Accept clipboard from viewers", "Send clipboard to viewers", "Send primary selection to viewers". I'm setting up a VNC server on a Debian 10 machine that has Gnome environment. A zillion articles can be found on Internet, they have the same global path: Install Gnome and GDM Install TigerVNC Setup the VNC Server: password + startup Run the server Connect from the client One of them is this TeknoTut how-to I took inspiration from. The tutorial is good, but I needed to complete it with extra steps in order to make it work: If the current user has already a running Gnome session, I just get the "grey screen with the checkboxes". It is good to know that: The "grey screen" is due to the "xsetroot" command that can be seen in some howtos The checkboxes are due to the "vncconfig -iconic &" command