Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I am in charge of a large number of Windows servers. Recently, many have been reporting hard drive errors with event codes 11 and 55. CHKDSK indicates that the drives are fine most of the time. What other diagnostic tools could I use to more accurately detect hard drive failures? Could these Windows events be false positives? I have already evaluated S.M.A.R.T., and it seems to have significant sensitivity and specificity issues.

share|improve this question

closed as primarily opinion-based by Ward, Dave M, Scott Pack, Jenny D, mdpc Jun 27 '13 at 1:13

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise.If this question can be reworded to fit the rules in the help center, please edit the question.

What "significant sensitivity and specificity issues"? It's what everyone uses –  Chopper3 Jun 25 '13 at 17:56
Thanks for the quick response! Often, failed drives give exactly zero indication, and indicators give a very weak suggestion that failure is possible in the near to distant future. Hence, significant sensitivity and specificity issues. See reference #2 here: en.wikipedia.org/wiki/S.M.A.R.T –  Francis Jun 25 '13 at 18:01
I don't need a link to that page, seriously - it's what everyone uses, whether you feel you have to set custom thresholds or whatever that's pretty much your only option. –  Chopper3 Jun 25 '13 at 18:08
Remember that chkdsk only checks the logical structure of the disk (the filesystem) unless you're specifying /r -- and frankly if you get to the point of using chkdsk /r (and it finds bad sectors) you should consider your drive dead. –  voretaq7 Jun 25 '13 at 18:31
There is no oracular way to devise whether a drive will fail. SMART attempts to predict it using available information, but sometimes they just die. –  Falcon Momot Jun 25 '13 at 18:32
add comment

2 Answers

up vote 6 down vote accepted

You detect hard drive failures by monitoring your RAID controller (or software RAID status) for drive failures.
When a drive fails, you replace it as quickly as possible.

Anything else is a proxy for predicting failure (which is useful, though not as critical with RAID).

At the moment there is no better tool than SMART for predicting failure (the very article you reference - which is still the definitive work 6 years later - shows a definite correlation between certain SMART errors and drive mortality).

SMART based monitoring suffers from a high "False Negative" rate, but positive predictions of failure can be regarded as extremely reliable (and false negatives are accounted for, again, by RAID).

share|improve this answer
Vote++. This is a basic systems management task. Have hardware. Have failures. Therefore, monitor your hardware status. Simples. –  Simon Catlin Jun 25 '13 at 19:29
Thanks! This is quite helpful. Would a SAS controller have similar functionality available? Also, what types of software could monitor the controllers? –  Francis Jun 25 '13 at 20:02
SAS (in fact most modern SCSI) controllers have similar functionality to SMART - I believe smartd can talk to the drives. RAID controllers usually have their own monitoring agents (or report via IPMI) - you would have to check your specific monitoring solution for integration instructions. (I know on Dell systems if you install OpenManage you get pretty extensive monitoring/reporting capabilities. IBM and HP have equivalent software available.) –  voretaq7 Jun 25 '13 at 21:56
add comment

Depending on the server manufacturer there is probably a tool or tools made to monitor the hardware from a central console. Dell uses Openmanage which will generate alerts for problems that are hardware related. HP and IBM have similar tools.

share|improve this answer
add comment

Not the answer you're looking for? Browse other questions tagged or ask your own question.