Yea, we had the exact same problem. VM HA is broken in 4.9 - at least under KVM
We use this Pull Request in our environment to fix it. https://github.com/apache/cloudstack/pull/2474 However, as stated in the PR, enable libvirt (edit /etc/libvirt/qemu.conf and change and comment: lock_manager = "lockd") https://libvirt.org/locking-lockd.html -----Original Message----- From: Parth Patel [mailto:[email protected]] Sent: Saturday, February 24, 2018 10:54 PM To: [email protected] Subject: Re: HA-enabled VM not starting or migrating to another host if current host goes down Hi Simon, I'm using KVM hypervisor. On Sun 25 Feb, 2018, 01:41 Simon Weller, <[email protected]> wrote: > Which hypervisor are you using? > > Simon Weller/615-312-6068 > > -----Original Message----- > From: Parth Patel [[email protected]] > Received: Saturday, 24 Feb 2018, 11:33AM > To: [email protected] [[email protected]] > Subject: HA-enabled VM not starting or migrating to another host if > current host goes down > > Hi, > > I am developing an enterprise-level cloud infrastructure currently > using Cloudstack 4.9. My head of department wished to check a specific > failsafe scenario. It is as follows: > > A highly available VM whose continuous execution even in event of some > storage or connection error, should remain running or migrate to > another host automatically if anything were to happen to the current > host. I tried removing the LAN cable from the current host after an > HA-enabled VM was executing on it, but the management server would not > auto-start the VM on another host. It kept printing error messages > such as "Communication failure. Host 5 timed out due to even > PingTimeout" in management server logs. I have manually set the ping > timeout duration to 30 seconds and its multiplier value to 1. > > Any suggestions as to what extra configuration is needed to make > Cloudstack start the VM on another host or migrate it? > > Just to be clear, the host on which the HA-enabled VM is running does > not have its primary and secondary storage added to the management > server, so the primary and secondary storage disks on NFS shares are > already available to the management server if it decides to start the > instance on another suitable host (of which there are 2). > > I tried searching some feature like this in Cloudstack administration > docs but could not find anything fitting to this scenario. > > Regards, > Parth Patel >
