Alert when a SMART counter moves, not when it says PASSED
the hard drive in my NAS says PASSED with 2,184 reported uncorrectable errors on it. an alert on the overall status would never fire, and an alert on “187 > 0” would fire forever. so my status page alerts when a counter moves off the value it had when I decided to keep the disk.
the baselines, keyed by model so a disk that comes back as sdc instead of sdb keeps its history:
| Disk | Baseline |
|---|---|
| Seagate ST9320423AS | 5=26, 187=2184, 197=0, 198=0 |
| KLEVV SSD | 197=0, 198=0 (no attribute 5) |
any change is critical and goes in the alert card. so is a failed long self-test. a long test older than 90 days is a warning on the server line.
it reads every disk with:
smartctl -n standby -i -A -l selftest /dev/sdb
-n standby skips a disk that’s asleep instead of spinning it up for a status page. the last reading is kept in a cache file and shown with its date, so a sleeping disk shows its last numbers and how old they are, not a blank.
the rule from the disk test still stands: the Seagate holds second copies only. this just tells me when that disk’s history starts moving again.
see also: SMART said PASSED with 2,184 errors, a monitoring key that can only run one script