How many pending sectors is too many?

[Ryan Beall]

I recently had a drive start popping bad sectors and the count got all the way to 80. A day later I had a second drive start popping sectors so I rushed out and replaced the drive with 80 bad sectors just in case. Now the second drive that started failing has 40 bad sectors and has stayed constant at 40 bad sectors for a month now. I've tried to offline the drives and write zeros on the sectors but can't seem to get the pending message to go away. Might not have the process down correctly. Either way, it seemed like all of the forum post tutorials on similar things all are correcting 1 pending sector. Is 40-80 bad sectors a no brainer to replace the drive and RMA it? Cant really find the number or spare sectors that are reasonable to be able to pull out of pending?

 

[Stux]

Yes. Replace it.

It's only an issue worth considering if the drive is out of warranty.

 

[Robert Trevellyan]

I tend to agree with @Stux, 40 is a bit high.

I've tried to offline the drives and write zeros on the sectors but can't seem to get the pending message to go away.

You could try this:

1. Shut down FreeNAS.
2. If you have another system available, connect the drive to that.
3. If not, disconnect all the other drives from the existing system.
4. Fire up an Ubuntu Live DVD and run badblocks -svw on the drive.

If this doesn't remap the pending sectors, or if it reports errors after the first write/read pass combination, replace the drive anyway.

 

[danb35]

There isn't really a firm answer on how many pending sectors is too many. I ran a drive with 60-some for a few months, until it started failing SMART tests. But in general, I'd lean toward thinking 40 is too many. If you want to stretch it, though, try what @Robert Trevellyan suggests. If the number doesn't increase after running badblocks, and badblocks itself doesn't report any errors, you could keep using the drive, but keep a close eye on it.

Personally, if it's under warranty, I'd RMA it.

 

[joeschmuck]

In my opinion if you have a drive under warranty, any count in the Pending Sector is a bad thing. While it is true, it can fix itself, meaning the pending sector could go away without increasing the ID 5 value (which should be 0), but my opinion is if I saw a count of 5 or greater in the Pending Sector value, I'd RMA the drive. If you have any value in ID 5, RMA the drive. If you are no longer covered by the RMA, do as @Robert Trevellyan suggested.

 

[Robert Trevellyan]

the pending sector could go away without increasing the ID 5 value

I think it might be a bit misleading to refer to that as "fix itself". From what I've read and experienced, it just indicates that the drive wrote to the sector(s) without reporting an error, and tells you nothing about whether it will be able to read what was written. To my mind, it's more desirable to see ID 5 and ID 196 increment when ID 197 decrements. The main reason I just recently returned a 2TB WD Red was precisely this, i.e. the drive repeatedly failed to reallocate questionable sectors when writing to them, yes it was unable to read from them. It's also why I think a full four-pass destructive badblocks -svw is the right tool for the job.

 

[joeschmuck]

I think it might be a bit misleading to refer to that as "fix itself".

I completely agree but I didn't feel like typing out the entire way it works. You probably know that I can get long in the tooth.

It's also why I think a full four-pass destructive badblocks -svw is the right tool for the job.

You should add the switch "-p 4" so the command line looks like badblocks -svw -p 4 (I like to separate some switch commands) and that will run it four times, well it will run it four times without errors. I've never had an error on a drive yet for running badblocks so I'm not sure what a small failure/fix runs like.

 

[Robert Trevellyan]

I completely agree

My goal was not to correct you but to avoid misleading others.

add the switch "-p 4"

When I run badblocks -svw it does four write passes, each followed by a read pass.

 

[joeschmuck]

My goal was not to correct you but to avoid misleading others.

That's cool, I need correcting often, just ask my wife.

When I run badblocks -svw it does four write passes, each followed by a read pass.

Well that is actually four different test patterns, each test pattern designed to verify the different bit within the electronics and magnetic patterns on the platter(s). I guess you could consider it four separate erase/write/read passes but it would only be considered one pass of badblocks.

 

[Stux]

Here's the thing....

If it takes 3 days for one pass and 12 for 4....

And you're burning in a disk to replace a failed z2 member...

Would you rather replace the member after zero, one or four badblocks passes?

I'd do it after one, to check for infant mortality, but I wouldn't want to wait the full two weeks for four passes.

 

[joeschmuck]

@OP
I'm chaining my opinion on what you should do based on lack of information provided... Have you run a SMART Long test? Does it fail the Extended Read test? I would expect that it will. If the drive is under warranty then RMA it. If it's out of warranty then you can run badblocks on just the section that failed (+/- 100,000) and then run a long SMART test and see if there are any further read issues, repeat the badblocks for the failed sectors, eventually when all looks good then run badblocks for a single pass. It may take a bit longer than just running badblocks alone but if you find other problems early in the testing then I'd cut my losses and stop testing, it's not worth having a questionable drive in your server.

 

[joeschmuck]

I'd do it after one, to check for infant mortality, but I wouldn't want to wait the full two weeks for four passes.

That wasn't my intent. My intent was to state that there are four different patterns being tested, it's not four complete tests.