Hello,
I am running PVE Cluster, but without any shared storage yet, since the nodes are located in different data centers around the globe. Mostly the cluster works as a central management solution and useful GUI tool for VM non-live migrations.
I am planning to install a separate node just for backup purposes of all other nodes and can't pick the best shared storage type/technological setup for my case. My considerations:
1) Since I'm using ZFS, I plan to use pvezsync as level-1 backup with snapshots of all VMs for the last few weeks. This works and is actually the best solution for me without even a need for shared storage at all, but has a problem in my special case. Since logically both the original host and backup host have to keep exactly the same all incremental snapshots stored, I'll run out of space on my original hosts very fast. The backup host is loaded with a bunch of HDDs and storage is not a problem there, but on original hosts, I am using NVMe's. Due to the nature of the data (it's pretty dynamic, weekly change is about 60%, but the amount is permanent - around 400GB per host) - keeping incremental snapshots for more than a few weeks will cost on NVMe simply too much.
2) So as the level-2 I plan to do regular backups weekly to shared storage. And now my question is what is the best practice/suggested method to configure a separate backup node with the only purpose of providing shared storage? It seems to me Ceph and Guster to be overkill for such use... I am thinking about launching there small debian KVM with docker-samba container with a simple CIFS share. Is there a better way I am overseeing atm?
Thanks for all your ideas shared
I am running PVE Cluster, but without any shared storage yet, since the nodes are located in different data centers around the globe. Mostly the cluster works as a central management solution and useful GUI tool for VM non-live migrations.
I am planning to install a separate node just for backup purposes of all other nodes and can't pick the best shared storage type/technological setup for my case. My considerations:
1) Since I'm using ZFS, I plan to use pvezsync as level-1 backup with snapshots of all VMs for the last few weeks. This works and is actually the best solution for me without even a need for shared storage at all, but has a problem in my special case. Since logically both the original host and backup host have to keep exactly the same all incremental snapshots stored, I'll run out of space on my original hosts very fast. The backup host is loaded with a bunch of HDDs and storage is not a problem there, but on original hosts, I am using NVMe's. Due to the nature of the data (it's pretty dynamic, weekly change is about 60%, but the amount is permanent - around 400GB per host) - keeping incremental snapshots for more than a few weeks will cost on NVMe simply too much.
2) So as the level-2 I plan to do regular backups weekly to shared storage. And now my question is what is the best practice/suggested method to configure a separate backup node with the only purpose of providing shared storage? It seems to me Ceph and Guster to be overkill for such use... I am thinking about launching there small debian KVM with docker-samba container with a simple CIFS share. Is there a better way I am overseeing atm?
Thanks for all your ideas shared
