MHDD Failure Threshold

[Mainstay]

Hi All,

This is likely covered somewhere on the forums but I cannot find the right search pattern to yield a result.

I am curious, in using MHDD (remap off, loop/repair off, standard scan), what do you consider your threshold for declaring a drive as "failed".

I am also curious as to how you define "failed".

On business-critical system I consider any slow access rates in the >500 ms range to be "failed". I clone to a new drive and move on. Any UNC's are an automatic fail. <500ms is a bit of judgment call (based on age of drive, performance of system, total cost of repairs (i.e., cost-adverse clients), total cost of fixing the situation should the drive fail in the field, etc.).

For non-critical systems I use a little more discretion. So long as a backup system exists I don't mind a few slow sectors (although I give fair warning to the client that it is showing signs of deterioration).

How do you judge drives based off of MHDD?

 

[sapphirescales]

When in doubt, replace. How much is a hard drive nowadays? I ask because I barely know anymore. SSD's are in, HDD's are out except for large data storage. And when it comes to data, is $40 really going to break you? I'm to the point where I'm replacing hard drives that are just fine just for the increased reliability and speed of SSD's.

 

[GTP]

How much is a hard drive nowadays?

This^^

I am also curious as to how you define "failed".

My simple rule is "any doubt? Toss it out!"

 

[lcoughey]

Some drives will have the occasional slow seek, so don't worry about anything under the <500ms threshold (unless there are a lot). The >500ms, UNC, AMNF and other errors are solid reasons to transition the drive from a usable drive to a drive that is good only for parts.

 

[labtech]

If I see entries for <500ms and >500ms, then I would categorize this as "on its way to failing". If mission critical, I suggest replacing.
Errors of any kind, categorized as "failed".

 

[Mainstay]

Thank you all.

I am completely on board with the "if in doubt" concept and appreciate that drives are cheap.

I was more interested in the metrics / rule of thumb that you have for your shop and drives, like those shared by @lcoughey and @labtech.

 

[Krynn72]

Any single error is cause to replace for me, and many slow seeks is as well. We often see these computers in for complaints of "running slow" so i'm not going to give them back a computer and tell them its "running nice and fast now" if 1/3 of their drive is reading at >150ms.