Comment 89 for bug 438136

Revision history for this message
Jean-Louis (jean-louis) wrote :

Hi, sorry for my bad english.

I would reassure some persons... the only catastrophic event is when palimpsest say BAD STATUS.

BAD STATUS is the only secure flag, because it is the response from SMART RETURN STATUS command and is the only official method for manifacturer to advise: imminent "disaster".

Remaining flags is set by libatasmart comparing normalized/worst value with threshold value.

Normalized, worst, current and threshold values is vendor dependent and the D1h command (read threshold) is obsolete... so this values isn't standard and some manufacturer may change their meaning.

libatasmart can only predict problems with heuristics and this isn't easy.

BAD_SECTOR_MANY is due to unlucky heuristic: is used "sector_threshold = u64log2(d->size/512);" this means that for an HDD of 1TB = 2^40Byte -> sector_threshold = 31sectors and obviously is too few

The red "ID, Attribute, Assessment and Value" in Palinsest is due to another problem: in function verify_sectors, is raised a->warn also for only 1 reallocated sector... this is a good thing, but the alarm is excessive.

I think that is a good idea to have more granular visual alarm. for example:

RED color -> assessment: FAILED (comparing the threshold, failed now)
ORANGE color -> assessment: WARNING (comparing the threshold, failed in past)
YELLOW color -> assessment: CAUTION (without comparing the threshold)
GREEN color -> assessment: GOOD

I would add a proposed patch for little fix libatasmart, obviously I don't have now a good heuristics for bad sectors ;)