Zfs special device "effect" comes on garbage collection, prune, verify runs.
really slow restore from HDD pool
- Thread starter carl0s
- Start date
I know, you did say so earlier, but prior to that, everybody else is telling me that my HDDs are the reason this is performing a bit below my hopes.
Now, when I look at the data overall:
inside the guest, there is 550GB used on the two virtual disks.
A full backup is taking up 175GiB on PBS (big texty SQL databases..)
Restoring 550GB in 19 minutes is pretty good I suppose.
Anyway I will carry on testing other filesystems etc. It still seems odd to me that the network is barely being used beyond 1.25Gbps (175GiB read in 19 minutes)
Last edited:
I believe I am being held up by pbs-restore being single threaded. I see it pegged at 100% on the PVE host.. :-/
Bit of a necro, but was seeking info on the same issue and I'm having similar results. Backing up is fast at 450-500MiB/s over a 10gbe line to a 18TB mirror with special device. However, restoring is far slower at 85-128MiB/s over the same line.
CPU usage is at 6.x% which, after discounting hyperthreading 'cores', results in a 1-core 100% utilization. IOwait is <5%, disks aren't busy.
Very happy with the backing up performance and deduplication, however, PBS needs optimizing to be a competitive solution when it comes to restoring.
Last edited: