Hi,
I know this has been discussed a bit in a previous thread, such as:
https://forum.proxmox.com/threads/dirty-new-bitmaps.81754/page-3
but I wanted to see if there has been any movement in the opinion of allowing administrators to force bitmap persistence on stopped VMs.
As Krzysztof mentioned here:
https://forum.proxmox.com/threads/dirty-new-bitmaps.81754/post-513324
The issue gets worse when people have very large virtual disks attached to VMs. The problem is exacerbated by the fact that only a single backup per host can be run at a time.
As Proxmox continues down the path to be a VMware replacement, this problem starts to become a showstopper, since vSphere, and respective backup products, can use the CBT (change block tracking) data (just like the dirty bitmap info in QEMU) to determine whether any data has changed, including for stopped/powered-off VMs. There is no need to read every byte of data in vSphere, so I suspect you are going to see a lot of people shaking their heads as to why this is such an issue when they try Proxmox.
Allowing administrators to determine their own fate should be an acceptable risk here - especially in well controlled environments where there is no way any changes could have been made to a virtual disk outside of Proxmox. If the Proxmox folks don't like this idea, maybe make this as a permanently experimental feature that can be enabled in the cluster config?
As a service provider, we could easily run into situations where users shut down numerous VMs that have large virtual disks, which will bombard our networked storage with reads that are completely unnecessary as well as using unnecessary bandwidth on the NICs on both the hosts as well as the networked storage devices. So, the question is also self-serving, but I can see that VMware migrations are going to see this as a major issue too (we also perform VMware migrations to Proxmox).
Would love to hear today's view on this, now that it has been a couple years since the above post.
Eric
I know this has been discussed a bit in a previous thread, such as:
https://forum.proxmox.com/threads/dirty-new-bitmaps.81754/page-3
but I wanted to see if there has been any movement in the opinion of allowing administrators to force bitmap persistence on stopped VMs.
As Krzysztof mentioned here:
https://forum.proxmox.com/threads/dirty-new-bitmaps.81754/post-513324
The issue gets worse when people have very large virtual disks attached to VMs. The problem is exacerbated by the fact that only a single backup per host can be run at a time.
As Proxmox continues down the path to be a VMware replacement, this problem starts to become a showstopper, since vSphere, and respective backup products, can use the CBT (change block tracking) data (just like the dirty bitmap info in QEMU) to determine whether any data has changed, including for stopped/powered-off VMs. There is no need to read every byte of data in vSphere, so I suspect you are going to see a lot of people shaking their heads as to why this is such an issue when they try Proxmox.
Allowing administrators to determine their own fate should be an acceptable risk here - especially in well controlled environments where there is no way any changes could have been made to a virtual disk outside of Proxmox. If the Proxmox folks don't like this idea, maybe make this as a permanently experimental feature that can be enabled in the cluster config?
As a service provider, we could easily run into situations where users shut down numerous VMs that have large virtual disks, which will bombard our networked storage with reads that are completely unnecessary as well as using unnecessary bandwidth on the NICs on both the hosts as well as the networked storage devices. So, the question is also self-serving, but I can see that VMware migrations are going to see this as a major issue too (we also perform VMware migrations to Proxmox).
Would love to hear today's view on this, now that it has been a couple years since the above post.
Eric
