Hello Everyone,
we inherited a client with a very non ideal backup concept. The backups are done by a PBS VM that lives on a PVE Host. The PBS host has a permanent datastore on a local mountpoint that mounts an NFS share on a synology NAS over 10G Link to a RAID5 Array of HDDs. This is first backup-tier and works mostly reliably. The second backup-tier is set up as follows:
- 5 HDDs
- Permanent Datastore on each HDD
- HDDs are connected to NAS via external USB on each day of the week
- NAS automagically creates statically named CIFS-share (i.e. backup-mon, backup-tue and so on)
- These CIFS shares are mounted by autofs at backup time, and unmounted when idle for 5 minutes
This works well, assuming there is not a single hickup in the cifs connection. Backup times are much much faster than the previously used Synology Hyperbackup, and storage is also used much more efficiently. However, any slight hickup causes the backup to fail with the following message:
As to why we cannot use NFS: Synology recognizes each disk individually and creates automagical, correctly named CIFS shares, however it always locally mounts the usb disk to the same path. And my limited knowledge about NFS tells me that the way it works is, that it basically exports that path, and doesnt label it based on anything.
Now I thought about this for a little bit and think a combination of USB-passtrough from the PVE Host and using the "removable Datastore" feature could fix this backup.
My idea was the following:
- label each external HDDs filesystem (i.e. backup-mon)
- automount the HDD to a directory corresponding to the current day (i.e. /mnt/backup-mon) based on the label
- add removable datastore to pve that points to device (i.e. /dev/sdb) and path (i.e. /mnt/backup-mon)
However I dont know if using the same device identifier (sdb) for different datastores (/mnt/backup-tue) will work.
If anyone has set up something similar, or if you have any other suggestions for me, I would be very grateful.
Have a nice day and a nice weekend.
we inherited a client with a very non ideal backup concept. The backups are done by a PBS VM that lives on a PVE Host. The PBS host has a permanent datastore on a local mountpoint that mounts an NFS share on a synology NAS over 10G Link to a RAID5 Array of HDDs. This is first backup-tier and works mostly reliably. The second backup-tier is set up as follows:
- 5 HDDs
- Permanent Datastore on each HDD
- HDDs are connected to NAS via external USB on each day of the week
- NAS automagically creates statically named CIFS-share (i.e. backup-mon, backup-tue and so on)
- These CIFS shares are mounted by autofs at backup time, and unmounted when idle for 5 minutes
This works well, assuming there is not a single hickup in the cifs connection. Backup times are much much faster than the previously used Synology Hyperbackup, and storage is also used much more efficiently. However, any slight hickup causes the backup to fail with the following message:
Backup of VM 311 failed - backup write data failed: command error: write_data upload error: pipelined request failed: inserting chunk on store 'Job3_Donnerstag' failed for cb734c388972948eb7741b82de15dc57ac3a86096584a40ad9eb06d05beea647 - write failed: Bad address (os error 14)As to why we cannot use NFS: Synology recognizes each disk individually and creates automagical, correctly named CIFS shares, however it always locally mounts the usb disk to the same path. And my limited knowledge about NFS tells me that the way it works is, that it basically exports that path, and doesnt label it based on anything.
Now I thought about this for a little bit and think a combination of USB-passtrough from the PVE Host and using the "removable Datastore" feature could fix this backup.
My idea was the following:
- label each external HDDs filesystem (i.e. backup-mon)
- automount the HDD to a directory corresponding to the current day (i.e. /mnt/backup-mon) based on the label
- add removable datastore to pve that points to device (i.e. /dev/sdb) and path (i.e. /mnt/backup-mon)
However I dont know if using the same device identifier (sdb) for different datastores (/mnt/backup-tue) will work.
If anyone has set up something similar, or if you have any other suggestions for me, I would be very grateful.
Have a nice day and a nice weekend.