I start with some information about my setup.
Proxmox 8.4.14
zfs raidz2 7x20TB seagate x20 20TB HDDs
No errors, and everything is pretty empty
A mirrored special device
I have a proxmox zfs storage created from a dataset on this pool. I then made a virtual disk from that storage that I gave to a vm to run smb on.
The vm’s are running Arch.
I’ve been having an issue on this build where transfers intermitently stop, for a minute or so then they will return to normal speeds for a similar amount of time. This has been going on for about a year. These transfers are often inter dataset so it will be ~150MiB/s normally and average out to ~45/s with the downtime.
Reading from the pool is fine, I can copy to my desktop at the full speed of my 2.5gib nic. ~250MiB/s
Downloading torrents doesn’t seem to have any problems. Maybe because it doesn’t exceed ~40MiB/s?
I’ve only seen a couple of people that say they have a similar situation, but one of them had smr drives, and the other was using dedupe and I’m not using either of these. The reason I’m making this post is because it’s stopped recovering and now doing “fast” writes seems to be completely locking up the virtual disks on the vm.
So on the same vm if I copy something to Disk A, then when it stops, the torrent which is reading off of Disk B will also stop. And during this time the cpu and disk wont be doing anytihng. Transfers between local disks will cause this so I don’t think it’s a samba or network issue.
This also affects both vitual machines that I have hosting my shares the same way. There is also nothing in dmesg when this happens.
It feels like it’s filling up some write cache, then needs to flush it all to the disk, but now the copy is just never resuming. Sorry if this is rambling, I’m just at a loss. Thank you in advance.