Hello,
I have a big pool of HDD storage in RAID 10 alongside NVMe Drives on my proxmox cluster.
The VM Data Drives are stored on the NVMe storage while backups are taken on the Local HDD Storage which is shared over NFS. The NFS share is in the same building as the Hypervisors.
In addition to the local HDD backup, I am looking into a mechanism to keep a copy of the backups at a remote (offshore) location which has around 70ms of latency and is running PBS accessible over public networking.
I have 10Gbit/s ports on each side so bandwidth isn't too much an issue. However, whenever the backup is running for a VM, it essentially becomes unresponsive. I did some digging and came to the understanding that the backup store is used as a buffer when the backup task is running for consistency reasons (correct me if I'm wrong) which would explain the unresponsiveness of the VM during backup.
I've considered using rsync from the HDD Pool backups but that would not be too practical since we have around 30TB of backups (considering that the data will be read from the HDD RAID which would be a bottleneck hence very time consuming - initial sync is doable but backing up all data every single time is not practical) and every rsync would essentially be non-incremental since the modified/created date on the vzdump file would be newer after the local backup is run.
Exact VM consistency is not a huge concern since this offsite backup is for worst-case disaster recovery scenarios only.
What are my options here?
Thanks
I have a big pool of HDD storage in RAID 10 alongside NVMe Drives on my proxmox cluster.
The VM Data Drives are stored on the NVMe storage while backups are taken on the Local HDD Storage which is shared over NFS. The NFS share is in the same building as the Hypervisors.
In addition to the local HDD backup, I am looking into a mechanism to keep a copy of the backups at a remote (offshore) location which has around 70ms of latency and is running PBS accessible over public networking.
I have 10Gbit/s ports on each side so bandwidth isn't too much an issue. However, whenever the backup is running for a VM, it essentially becomes unresponsive. I did some digging and came to the understanding that the backup store is used as a buffer when the backup task is running for consistency reasons (correct me if I'm wrong) which would explain the unresponsiveness of the VM during backup.
I've considered using rsync from the HDD Pool backups but that would not be too practical since we have around 30TB of backups (considering that the data will be read from the HDD RAID which would be a bottleneck hence very time consuming - initial sync is doable but backing up all data every single time is not practical) and every rsync would essentially be non-incremental since the modified/created date on the vzdump file would be newer after the local backup is run.
Exact VM consistency is not a huge concern since this offsite backup is for worst-case disaster recovery scenarios only.
What are my options here?
Thanks
Last edited: