High IO Delay? 50-80%

[shawn999]

Hello,

I think I have some problems with nvme drive speed and io delay and maybe a lot of things.
These are my specs:

AMD Ryzen 5 1600 6 cores 12 threads
Gigabyte aorus b450 elite v2
Crucial Ballistix BL2K8G32C16U4B 3200 MHz 32gb (4x8gb)
Samsung evo 980 1TB in m2a slot
Another Samsung evo 980 1TB in a pcie 3.0 adapter because the m2b slot is only 3.0 x2
Kingston a400 120gb as boot drive

I have my 2 980's in a striped zfs because I don't care what happens to anything because I have backups and use mergerfs with cloud storage. I have a feeling that my drives or something else is not working like they should. My IO delay is usually 50 to 80 % when I'm building mkv's from blu-rays in makemkv. I already returned the other 2 980's because I thought they were broken. But with 2 new ones I still have the same problems.

pool:
    version: 5000
    name: 'pool'
    state: 0
    txg: 4
    pool_guid: 17648312708667109412
    errata: 0
    hostid: 2615562297
    hostname: 'deathstar'
    com.delphix:has_per_vdev_zaps
    vdev_children: 2
    vdev_tree:
        type: 'root'
        id: 0
        guid: 17648312708667109412
        create_txg: 4
        children[0]:
            type: 'disk'
            id: 0
            guid: 3061452878188404099
            path: '/dev/nvme0n1p1'
            devid: 'nvme-Samsung_SSD_980_1TB_S649NX0T400287D-part1'
            phys_path: 'pci-0000:01:00.0-nvme-1'
            whole_disk: 1
            metaslab_array: 139
            metaslab_shift: 33
            ashift: 12
            asize: 1000189984768
            is_log: 0
            create_txg: 4
            com.delphix:vdev_zap_leaf: 66
            com.delphix:vdev_zap_top: 67
        children[1]:
            type: 'disk'
            id: 1
            guid: 12477372386210996944
            path: '/dev/nvme1n1p1'
            devid: 'nvme-Samsung_SSD_980_1TB_S649NX0T400291H-part1'
            phys_path: 'pci-0000:07:00.0-nvme-1'
            whole_disk: 1
            metaslab_array: 128
            metaslab_shift: 33
            ashift: 12
            asize: 1000189984768
            is_log: 0
            create_txg: 4
            com.delphix:vdev_zap_leaf: 68
            com.delphix:vdev_zap_top: 69
    features_for_read:
        com.delphix:hole_birth
        com.delphix:embedded_data

As for benchmarking, I really don't know how to benchmark these drives in linux. I tried looking it up but the settings and options you give it make huge differences.

I am running 1 lxc container with docker and 1 virtual machine with home assistant. The zfs pool has 1 dataset which is mounted to the lxc container.
Even when the drives (as far as I know) aren't doing anything the io delay is still 20-35%.



Also in a program called "Bazarr" I am getting a ton of "io 5 errors". It's a program that downloads subtitles for movies and tvshows.

Traceback (most recent call last):
  File "/app/bazarr/bin/bazarr/subtitles/utils.py", line 39, in get_video
    video = parse_video(path, hints=hints, providers=providers, dry_run=used_scene_name,
  File "/app/bazarr/bin/bazarr/../libs/subzero/video.py", line 61, in parse_video
    return scan_video(fn, hints=hints, dont_use_actual_file=dry_run, providers=providers,
  File "/app/bazarr/bin/bazarr/../libs/subliminal_patch/core.py", line 811, in scan_video
    video.hashes[\'opensubtitles\'] = osub_hash = osub_hash or hash_opensubtitles(hash_path)
  File "/app/bazarr/bin/bazarr/../libs/subliminal/utils.py", line 32, in hash_opensubtitles
    filebuffer = f.read(bytesize)
OSError: [Errno 5] I/O error

Maybe it's relevant just giving as much info as I can.

Please help me. I have been running proxmox on this server for a few months now and I can't figure it out. I spent hours in bios settings, zfs settings/config, I even tried lvm and ext4 directory instead of zfs.

I don't know.........