[Bug]: SMART attribute 198 flags WARNING on raw value alone, ignoring normalised value and drivedb absence

Welcome!

• I have read the Documentation
• I have checked the Common Issues Guide and my problem was not mentioned there.
• I have searched open and closed issues and my problem was not mentioned before.
• I have verified I am using the latest version available.

Component

Agent

Problem Description

Up front, so nobody wastes time reproducing it my way: on my hardware this does not currently surface, because smartctl --scan classifies the drive as scsi and it never reaches parseSmartForSata. I am reporting the logic with a real drive as a counter-example, not a broken dashboard. If that is too speculative to act on, please close it — no argument from me.

The check added in #2275 (shipped in v0.19.0, agent/smart.go):

if smartData.SmartStatus == "PASSED" && rawValue > 0 && (attr.ID == 5 || attr.ID == 197 || attr.ID == 198) {
    smartData.SmartStatus = "WARNING"
}

The goal from #2268 is right — "PASSED" really does hide developing damage, and reallocated and pending sectors are good signals. The concern is narrower: attribute 198 keyed on raw value alone, with no corroboration, on drives where smartmontools has no vendor table and is therefore only guessing what 198 means.

A concrete drive, an HPE-branded SATA SSD, firmware HPG2:

ID# ATTRIBUTE_NAME          FLAG     VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE
  5 Reallocated_Sector_Ct   0x0033   100   100   036    Pre-fail  Always       -       0
187 Reported_Uncorrect      0x0032   100   100   000    Old_age   Always       -       0
196 Reallocated_Event_Count 0x0033   100   100   036    Pre-fail  Always       -       0
197 Current_Pending_Sector  0x0012   100   100   000    Old_age   Always       -       0
198 Offline_Uncorrectable   0x0010   100   100   000    Old_age   Offline      -       224
199 UDMA_CRC_Error_Count    0x003e   100   100   000    Old_age   Always       -       0

SMART overall-health self-assessment test result: PASSED
Device is: Not in smartctl database 7.5/5706

Four independent reasons 224 is not a bad-sector count here:

1. Normalised value 100, worst 100, threshold 0. The drive itself reports this attribute as never having degraded. A real uncorrectable count drives the normalised value down.
2. Device is: Not in smartctl database. With no vendor table, smartctl falls back to the generic ATA name list, where 198 is conventionally Offline_Uncorrectable. The vendor never defined it that way — the name is an artefact of the fallback.
3. It is an SSD. Offline-uncorrectable describes an HDD offline surface scan.
4. No corroboration. 5, 187, 196, 197 and 199 all read 0. Real uncorrectable sectors do not appear alone; they leave a trail across those counters.

Flags are 0x0010 — Old_age, offline-collected, not pre-fail — which is also not how a drive advertises an attribute it wants you to act on.

Expected Behavior

Something that preserves #2268's intent without firing on vendor-specific counters. Options, roughly in order of how cheap they are:

• Require the normalised value to be at or below threshold (or meaningfully below 100) rather than treating any nonzero raw as a fault. That alone would exclude this drive.
• Require corroboration for 198 specifically — e.g. only warn if 5, 187 or 197 are also nonzero. 198 alone, with everything else clean, is far more often a vendor counter.
• Skip 198 when smartctl reports the drive is not in its database, since the attribute name is then a guess. smartctl -j exposes this.
• Treat 5 and 197 (well-standardised) differently from 198 (widely reused by vendors).

Happy to open a PR for whichever of these you would accept — I have the drive to test against, though as noted I would need to force the sat parser to exercise the path.

Steps to Reproduce

1. Attach a SATA SSD that smartmontools has no drivedb entry for and that reports a nonzero raw value on attribute 198 while remaining PASSED with 100/100 normalised.
2. Ensure it is parsed by parseSmartForSata (on my host --scan says -d scsi, so it is not — see #2295 — which is why I cannot demonstrate the badge directly).
3. The status flips PASSED → WARNING on the strength of the raw value alone.

Environment

• Agent 0.19.0, hub 0.19.0
• Proxmox VE 9.2.10, smartctl 7.5
• Drive: MK000480GWXFF (HPE-branded SATA SSD), firmware HPG2