HA cluster + ~Zero Downtime

[BloodyIron]

Hi Folks,

I'm trying to build a lab cluster with the goal of having a HA cluster which experiences (almost) zero downtime. My understanding is that the RAM needs to be synced between at least two nodes. From what I understand, proxmox has HA clustering, but it _sounds_ like it just boots the system back up when a node fails. I'm not certain if my understanding is correct here, but my goal is to make it so that the VM only has a few packets of loss while it migrates to another node, and doesn't lose any state information.

Has anyone had any success with this for proxmox?

 

[andydills]

In my (admittedly brief) experience with the HA, that's not possible.

Before another node will bring up the VM, it has to make 100% sure the original node is truly down, to prevent data corruption. Because the node must be truly down before a secondary node will bring up the VM, there is no way to recover state. You're basically dealing with a freshly booted VM on the secondary node.

I don't believe the HA has the goal of zero downtime, or anything close...I believe the HA implementation's goal is to enable the restoration of service to the affected VMs without waiting for human intervention.

 

[spirit]

Hi Folks,

I'm trying to build a lab cluster with the goal of having a HA cluster which experiences (almost) zero downtime. My understanding is that the RAM needs to be synced between at least two nodes. From what I understand, proxmox has HA clustering, but it _sounds_ like it just boots the system back up when a node fails. I'm not certain if my understanding is correct here, but my goal is to make it so that the VM only has a few packets of loss while it migrates to another node, and doesn't lose any state information.

Has anyone had any success with this for proxmox?

Hi what do you want is the kemari qemu project
http://wiki.qemu.org/Features/FaultTolerance

But It has never been released/finished.