From 15e09f49d28cb6312a95d304466bb79cb1a85c5d Mon Sep 17 00:00:00 2001
From: batsumaru <>
Date: Wed, 1 Jul 2026 16:55:53 +0900
Subject: Make the zpool check per-pool with capacity, scrub, and error status

Replace the single "all pools are healthy" summary line with one row
per pool (name, health, capacity) plus a scrub/errors row parsed from
`zpool status <pool>`. A pool that's never been scrubbed but has no
data errors still reports ok - lack of scrub history isn't itself a
failure, only actual reported errors are.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
---
 statuspage/checks/hw.sh | 31 ++++++++++++++++++++++---------
 1 file changed, 22 insertions(+), 9 deletions(-)

(limited to 'statuspage/checks/hw.sh')

diff --git a/statuspage/checks/hw.sh b/statuspage/checks/hw.sh
index bc0c669..259234d 100755
--- a/statuspage/checks/hw.sh
+++ b/statuspage/checks/hw.sh
@@ -1,6 +1,7 @@
 #!/bin/sh
 # checks/hw.sh
-# Hardware/host metrics: load avg, uptime, free mem, zroot usage, zpool health.
+# Hardware/host metrics: load avg, uptime, free mem, zroot usage, and per
+# pool health/capacity plus scrub/error status for every zpool on the box.
 set -eu
 
 DIR=$(dirname "$0")
@@ -31,12 +32,24 @@ else
 fi
 json_line "hardware" "zroot usage" "$DISK" "$DISK_STATUS"
 
-if ZPOOL=$(zpool status -x 2>&1); then
-    case "$ZPOOL" in
-        "all pools are healthy") ZPOOL_STATUS="ok" ;;
-        *) ZPOOL_STATUS="warn" ;;
+for pool in $(zpool list -H -o name 2>/dev/null); do
+    HEALTH=$(zpool list -H -o health "$pool" 2>/dev/null)
+    CAP=$(zpool list -H -o capacity "$pool" 2>/dev/null)
+    case "$HEALTH" in
+        ONLINE) POOL_STATUS="ok" ;;
+        DEGRADED) POOL_STATUS="warn" ;;
+        *) POOL_STATUS="down" ;;
     esac
-else
-    ZPOOL_STATUS="down"
-fi
-json_line "hardware" "zpool" "$ZPOOL" "$ZPOOL_STATUS"
+    json_line "hardware" "zpool $pool" "$HEALTH, ${CAP} used" "$POOL_STATUS"
+
+    STATUS_OUT=$(zpool status "$pool" 2>/dev/null) || STATUS_OUT=""
+    SCAN=$(printf '%s\n' "$STATUS_OUT" | sed -n 's/^[[:space:]]*scan: //p')
+    ERRORS=$(printf '%s\n' "$STATUS_OUT" | sed -n 's/^[[:space:]]*errors: //p')
+
+    case "$ERRORS" in
+        "No known data errors") ERR_STATUS="ok" ;;
+        "") ERR_STATUS="warn" ;;
+        *) ERR_STATUS="down" ;;
+    esac
+    json_line "hardware" "zpool $pool scrub" "${SCAN:-unknown}" "$ERR_STATUS"
+done
-- 
cgit v1.3