vSphere Availability

 View Only
  • 1.  VM restart sequence after host failure

    Posted Aug 10, 2012 07:10 PM

    Hi,

    We have a new installation with 2 HA hosts. Is there a way to configure dependecies that can be checked when the VMs are restarted after a failure?

    I have configured the "VM Restart Priority" in the cluster settings, but it doesnt seem to be as configurable as I expected, it only offers high, medium, low settings.

    We basically just want some sort of child-parent relationship where the child vm can't start until the parent is fully started.

    Thanks



  • 2.  RE: VM restart sequence after host failure

    Posted Aug 10, 2012 07:56 PM

    As far as I know the restart priority defined in the cluster settings it the only way to achieve this.



  • 3.  RE: VM restart sequence after host failure

    Posted Aug 10, 2012 08:01 PM

    Those are the HA settings.  Can I assume you need the restart order because some services rely on the other servers, for example a web service needing the database to be online first?



  • 4.  RE: VM restart sequence after host failure

    Posted Aug 10, 2012 08:16 PM

    Yes, it's a database, webserver, and an app server. I was just trying to find the best solution for a failure where some of the servers might already be started depending on whatever host failed. Maybe just keeping servers that have depencencies like that on the same host is the solution?



  • 5.  RE: VM restart sequence after host failure

    Posted Aug 10, 2012 10:46 PM

    Something outside of VMware, like a monitoring solution that can detect outages in your web app and can restart services based on certain conditions might do the trick.  Maybe vCenter Orchestrator but honestly am not familiar enough with that to say for sure.



  • 6.  RE: VM restart sequence after host failure

    Posted Aug 11, 2012 12:21 AM

    Assuming all 3 VMs are on the same server, using the high / medium / low may possibly work. In reality, HA just uses that as a start order to ensure priority VMs get a better chance to power on.

    If they aren't on the same host, you're back to square one. Outside of brute force rebooting of the VM(s), it is the application's responsibility to detect a failure at a lower tier (web > app > db) and retry or alert accordingly, the same as in a physical environment.



  • 7.  RE: VM restart sequence after host failure

    Posted Aug 13, 2012 08:20 AM

    Chris Wahl wrote:

    Assuming all 3 VMs are on the same server, using the high / medium / low may possibly work. In reality, HA just uses that as a start order to ensure priority VMs get a better chance to power on.

    If they aren't on the same host, you're back to square one. Outside of brute force rebooting of the VM(s), it is the application's responsibility to detect a failure at a lower tier (web > app > db) and retry or alert accordingly, the same as in a physical environment.

    Even if they are on the same host it is no guarantee that the VMs will be restarted in that order. We are talking "power-on attempts" here not a full boot of the virtual machine. So in other words if you have a database server it might take 3 minutes to fully start while the web server only takes 1 minute. The power-on attempt for either will literally take seconds. So after a second or two the next VM will be started, meaning that the web server will be ready to go after 1 minute and then will have to wait for 2 minutes for the DB to be ready.