zfs: tune hdds pool

modules/hardware.nix

@@ -21,4 +21,15 @@
 
   hardware.cpu.amd.updateMicrocode = true;
   hardware.enableRedistributableFirmware = true;
+
+  # HDD I/O tuning for torrent seeding workload (high-concurrency random reads).
+  # mq-deadline sorts requests into elevator sweeps, reducing seek distance.
+  # Aggressive deadlines (15s) let the scheduler accumulate more ops before dispatching,
+  # maximizing coalescence — latency is irrelevant since torrent peers tolerate 30-60s.
+  # fifo_batch=128 keeps sweeps long; writes_starved=16 heavily favors reads.
+  # 4 MiB readahead matches libtorrent piece extent affinity for sequential prefetch.
+  services.udev.extraRules = ''
+    ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="mq-deadline", ATTR{queue/read_ahead_kb}="4096", ATTR{queue/nr_requests}="512"
+    ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="1", ENV{ID_ATA_ROTATION_RATE_RPM}!="0", RUN+="${pkgs.bash}/bin/bash -c 'echo 15000 > /sys$devpath/queue/iosched/read_expire; echo 15000 > /sys$devpath/queue/iosched/write_expire; echo 128 > /sys$devpath/queue/iosched/fifo_batch; echo 16 > /sys$devpath/queue/iosched/writes_starved; echo 4096 > /sys$devpath/queue/max_sectors_kb 2>/dev/null || true'"
+  '';
 }

modules/zfs.nix

@@ -10,6 +10,21 @@
 
   boot.kernelParams = [
     "zfs.zfs_txg_timeout=120" # longer TXG open time = larger sequential writes
+
+    # vdev I/O scheduler: feed more concurrent reads to the block scheduler so
+    # mq-deadline has a larger pool of requests to sort and merge into elevator sweeps.
+    # Default async_read_max is 3 — far too few for effective coalescence.
+    # 32 was empirically optimal (64 overwhelmed the drives, 3 gave near-zero merges).
+    "zfs.zfs_vdev_async_read_max_active=32"
+    "zfs.zfs_vdev_async_read_min_active=4"
+
+    # Merge reads within 128 KiB of each other (default 32 KiB). On HDDs, reading a
+    # 128 KiB gap is far cheaper than a mechanical seek (~8 ms).
+    "zfs.zfs_vdev_read_gap_limit=131072"
+
+    # Allow ZFS to aggregate I/Os up to 4 MiB (default 1 MiB), matching the
+    # libtorrent piece extent size for larger sequential disk operations.
+    "zfs.zfs_vdev_aggregation_limit=4194304"
   ];
 
   boot.supportedFilesystems = [ "zfs" ];