Good morning,
I setup up my first lab environment using Proxmox. My cluster setup is made up of 4 nodes distributed to 2 different sites and a qdevice installed on a VM on a third site in order to obtain odd votes for a correct quorum.
Network is stretched between the two sites as well as the storage that is based on Hitachi GAD technology.
HA feature works fine, I tried failover scenario disabling network to the hosts on one site and the virtual machines have been restarted correctly on the remote site.
Also for the storage in a failure scenario on one site with all active paths to the local storage system failed, no disruption has been observed because standby paths to the remote storage system became active without data loss.
Every host has 4 paths torward the SAN Lun, 2 active I/O against the local storage in the local site and 2 standby I/O against the remote storage in the opposite site.
However, I have problem with the scenario involving the loss of all paths, both local and remote on a proxmox node, even on a single node.
The guest virtual machines remain without disk access in a frozen state. No protection mechanism is triggered by the HA feature. I was hoping for an autofence response from faulty node, so that guest VMs could reboot on the other Proxmox nodes with working storage access, but nothing happened. To resolve the issue, I had to manually reboot the host or restore correct storage access.
How can I implement an APD protection mechanism in a PVE cluster?
Thanks
Gianluca
I setup up my first lab environment using Proxmox. My cluster setup is made up of 4 nodes distributed to 2 different sites and a qdevice installed on a VM on a third site in order to obtain odd votes for a correct quorum.
Network is stretched between the two sites as well as the storage that is based on Hitachi GAD technology.
HA feature works fine, I tried failover scenario disabling network to the hosts on one site and the virtual machines have been restarted correctly on the remote site.
Also for the storage in a failure scenario on one site with all active paths to the local storage system failed, no disruption has been observed because standby paths to the remote storage system became active without data loss.
Every host has 4 paths torward the SAN Lun, 2 active I/O against the local storage in the local site and 2 standby I/O against the remote storage in the opposite site.
However, I have problem with the scenario involving the loss of all paths, both local and remote on a proxmox node, even on a single node.
The guest virtual machines remain without disk access in a frozen state. No protection mechanism is triggered by the HA feature. I was hoping for an autofence response from faulty node, so that guest VMs could reboot on the other Proxmox nodes with working storage access, but nothing happened. To resolve the issue, I had to manually reboot the host or restore correct storage access.
How can I implement an APD protection mechanism in a PVE cluster?
Thanks
Gianluca