Does my concept make sense?

[Hunduster]

Hello everyone,
I've recently switched to PVE and PBS and have been thinking about a backup concept that I'm not sure will work the way I think it will

I have a NAS with spinners and SSDs. I have integrated two datastores on the PBS:

• Datastore 1 on SSD
• Datastore 2 on HDD

I have a backup job in PVE that performs a backup to Datastore 1 every hour. The data sets are verified directly after creation on Datastore 1.
Under Datastore 2 there is a sync job that syncs all backups from Datastore 1 to Datastore 2 at 00:00 at night.
At 01:00 the Prune and GC jobs then run on both datastores.

The Prune jobs are set as follows:

Datastore 1



Datastore 2



At 02:00 an offsite PBS then goes and pulls the data from Datastore 2.

My thought or goal behind the concept is that the hourly backups run on the more power-efficient SSDs so that the spinners in the NAS can go into sleep mode.
Once a day, the daily, weekly and monthly backups should then be copied to the spinners for “long-term archiving”.

I am now wondering whether this makes sense and whether PBS understands the connection between the two datastores and whether the backup chains are consistent or whether I am fooling myself here, especially with regard to consistency and deduplication.

I currently had 3 warnings last night under the GC job of Datastore 1: WARN: warning: unable to access non-existent chunk.
This makes me worry that my concept is not as well thought out as I initially thought.

 

[Johannes S]

For me this setup looks ok. But I would expect bad Performance for the remote sync of the offsite pbs, verify and Garbage collection on the hdd datastore. Did you run some benchmarks?
What size do you have on your ssds and hdds? It might be worth to add a mirrored ssd Partition as a special device for Metadata of the hdd data.
I would also setup the remote pbs to sync from the ssds for better performance

 

[Hunduster]

For me this setup looks ok. But I would expect bad Performance for the remote sync of the offsite pbs, verify and Garbage collection on the hdd datastore. Did you run some benchmarks?
What size do you have on your ssds and hdds? It might be worth to add a mirrored ssd Partition as a special device for Metadata of the hdd data.
I would also setup the remote pbs to sync from the ssds for better performance

I only have 50 Mbit/s upload at my home, the remote sync to Offiste is not faster.
The SSDs run in RAID 5 with 1.5 TB. My HDDs in RAID 6 with ~20 TB

 

[Johannes S]

I only have 50 Mbit/s upload at my home, the remote sync to Offiste is not faster.
The SSDs run in RAID 5 with 1.5 TB. My HDDs in RAID 6 with ~20 TB

Ok then you are propably right that there is not much (if at all) to gain from a sync from the HDDs. I'm still curious whether garbage collection and verify jobs on the HDD datastore have sufficient performance.

Concerning the warning about the missing chunks: As far I know missing chunks should be resynced with the next backup or sync job. Did you encounter this warning again?