#!/bin/bash
# Pool capacity and the last scrub for the node exporter textfile collector.
# A scrub in progress or cancelled keeps the previous end time: the pool has no newer completed scrub.
set -euo pipefail

dir={{ zfs_host_textfile_dir }}
out=$dir/zfs_pool.prom
tmp=$(mktemp "$dir/.zfs_pool.XXXXXX")
trap 'rm -f "$tmp"' EXIT

# Value of a metric line of the previous run, 0 if there is none.
previous() {
  awk -v m="$1" -v p="$2" 'index($0, m "{pool=\"" p "\"}") == 1 {v = $2} END {print v + 0}' "$out" 2>/dev/null || echo 0
}

status=$(zpool status -p)
{
  echo '# TYPE zfs_pool_size_bytes gauge'
  echo '# TYPE zfs_pool_allocated_bytes gauge'
  echo '# TYPE zfs_pool_fragmentation_ratio gauge'
  zpool list -Hp -o name,size,allocated,fragmentation | while IFS=$'\t' read -r pool size alloc frag; do
    echo "zfs_pool_size_bytes{pool=\"$pool\"} $size"
    echo "zfs_pool_allocated_bytes{pool=\"$pool\"} $alloc"
    if [[ $frag =~ ^[0-9]+$ ]]; then
      echo "zfs_pool_fragmentation_ratio{pool=\"$pool\"} $(awk -v f="$frag" 'BEGIN {print f / 100}')"
    fi
  done

  echo '# TYPE zfs_pool_scrub_end_timestamp_seconds gauge'
  echo '# TYPE zfs_pool_scrub_errors gauge'
  zpool list -Hp -o name | while read -r pool; do
    scan=$(awk -v p="$pool" '$1 == "pool:" {cur = $2} cur == p && $1 == "scan:" {sub(/^[ \t]*scan:[ \t]*/, ""); print; exit}' <<< "$status")
    end=0
    errors=0
    if [[ $scan =~ ^scrub\ repaired\ .*\ with\ ([0-9]+)\ errors\ on\ (.+)$ ]]; then
      errors=${BASH_REMATCH[1]}
      end=$(date -d "${BASH_REMATCH[2]}" +%s)
    elif [[ $scan =~ ^scrub\ (in\ progress|canceled) ]]; then
      end=$(previous zfs_pool_scrub_end_timestamp_seconds "$pool")
      errors=$(previous zfs_pool_scrub_errors "$pool")
    fi
    echo "zfs_pool_scrub_end_timestamp_seconds{pool=\"$pool\"} $end"
    echo "zfs_pool_scrub_errors{pool=\"$pool\"} $errors"
  done
} > "$tmp"

chmod 0644 "$tmp"
mv "$tmp" "$out"
trap - EXIT