Hello everyone,
we have a couple of customers that run PVE on hardware that is not especially fast. Looking at the specs (Intel(R) Core(TM) i3-8109U CPU @ 3.00GHz, 32 GB RAM, 1TB SSD) you might think otherwise, but we have the following issue.
Some of the virtual machines are Debian based and come with multiple docker containers, separate nginx, mysql and apaches and other daemons. Some VMs are Windows based and run a number of services and programs. A regular boot/reboot of a single VM is fast (about 30 seconds) and the load is OK.
However, during a PVE backup in "stop" mode the virtual machines regularly cannot start all the daemons and services. The Debian logs show various errors (mostly kernel errors) of various daemons. Almost all of them have something to do with the load at this particular moment. It seems the PVE backup uses a lot of CPU time so the start of the VMs cannot be done gracefully. Right now we manage to stabilize the systems with various scripts that reboot the services before our customers get to work every day. But this is highly uncomfortable.
I know that there are other backup modes (suspend, snapshot) but their potential caveats (a backup might not be recoverable) are not acceptable for us.
A longer downtime however is not a problem. It's OK if the VMs are not reachable at night for a couple or hours.
So my question is: Could you please implement a backup option that shuts down all VMs included in the backup, performs the backup operation on all of them one after another and afterwards starts all VMs again? I believe that by doing this, the load on the system will be minimal and everything should boot up properly.
we have a couple of customers that run PVE on hardware that is not especially fast. Looking at the specs (Intel(R) Core(TM) i3-8109U CPU @ 3.00GHz, 32 GB RAM, 1TB SSD) you might think otherwise, but we have the following issue.
Some of the virtual machines are Debian based and come with multiple docker containers, separate nginx, mysql and apaches and other daemons. Some VMs are Windows based and run a number of services and programs. A regular boot/reboot of a single VM is fast (about 30 seconds) and the load is OK.
However, during a PVE backup in "stop" mode the virtual machines regularly cannot start all the daemons and services. The Debian logs show various errors (mostly kernel errors) of various daemons. Almost all of them have something to do with the load at this particular moment. It seems the PVE backup uses a lot of CPU time so the start of the VMs cannot be done gracefully. Right now we manage to stabilize the systems with various scripts that reboot the services before our customers get to work every day. But this is highly uncomfortable.
I know that there are other backup modes (suspend, snapshot) but their potential caveats (a backup might not be recoverable) are not acceptable for us.
A longer downtime however is not a problem. It's OK if the VMs are not reachable at night for a couple or hours.
So my question is: Could you please implement a backup option that shuts down all VMs included in the backup, performs the backup operation on all of them one after another and afterwards starts all VMs again? I believe that by doing this, the load on the system will be minimal and everything should boot up properly.
