Bad Block, Identify bad drive.

[Steve Simpson]

I'm getting continuous pages of the following error:
vm_fault: pager read error, pid xxxxx (nginx)
swap_pager: I/O error - pagein failed; blkno 1048907, size 4096, error 6

How can Identify which drive is the bad one? Will a zpool scrub mark that block as bad? Before I do that I would like to identify the bad drive so I can replace it.

After booting I get an Alert on the web interface which says to determine if the drive needs to be replaced yet the drive isn't identified.

Thanks for any help.

 

[Steve Simpson]

Now a different drive has died. Freenas says it has lost device ada1 and has removed the entry. So I replaced the drive with another. When I tried to add the new drive to the pool via the web interface my choice is "stripe" etc and there is no selection to add it to the ZFS pool. Any idea on how to add it to the existing zpool?

 

[Steve Simpson]

And now this (see screenshot) on another drive. Maybe I need a new motherboard, or replace the SATA cable.

 

[IanWorthington]

Have you run SMART tests? On ada3 as a minimum?

 

[Steve Simpson]

Thanks for your suggestion. I just added a smart test for all the drives. I can't tell if it has actually run or is now running. Where does the output go when it does run?

 

[dlavigne]

So I replaced the drive with another. When I tried to add the new drive to the pool via the web interface my choice is "stripe" etc and there is no selection to add it to the ZFS pool. Any idea on how to add it to the existing zpool?

You don't, the replacement process does that for you: http://doc.freenas.org/index.php/Volumes#Replacing_a_Failed_Drive.

 

[IanWorthington]

Thanks for your suggestion. I just added a smart test for all the drives. I can't tell if it has actually run or is now running. Where does the output go when it does run?

iirc that's a scheduled test which runs at the appointed time. I /believe/ you get an email if it fails but I've never seen one... In your case I'd run it at the command line, leave it for the recommended time, then you can issue another command to see the results.

See https://wiki.archlinux.org/index.php/S.M.A.R.T. for syntax

 

[Steve Simpson]

You don't, the replacement process does that for you: http://doc.freenas.org/index.php/Volumes#Replacing_a_Failed_Drive.

I was trying to replace the drive via the ZFS Volume Manager which seemed to be the right place for it. Right now the bad file server is down. Using my good master server I don't see a "Volume Status" tab. That would explain why I thought ZFS Volume Manager was the correct place to replace a drive. It seemed to be the only place. How do you get the Volume Status tab to show up? I put back the bad drive on the bad server to see if the tab shows up then but with the errors on ada1 it never even gets to the point where the web server is up.

It looks like I'll need to replace both ada1 and ada3 and start from scratch *again*.

 

[Steve Simpson]

I wound up replacing two drives and recreating the pool from scratch. All data lost. Copying everything from the master file server to the backup. It'll take 4 days or so. Again.

 

[IanWorthington]

I ran badblocks for over week on my drives before I trusted them with my data. And FOUR were bad.

 

[cyberjock]

I ran badblocks for over week on my drives before I trusted them with my data. And FOUR were bad.

Whoa! bad luck.. go buy a lottery ticket as luck has to swing back, right?

 

[IanWorthington]

Whoa! bad luck.. go buy a lottery ticket as luck has to swing back, right?

I did: I bought new drives and they were all good :)

 

[Ericloewe]

I ran badblocks for over week on my drives before I trusted them with my data. And FOUR were bad.

Out of how many? Sounds like someone viewed the "fragile" sticker as a challenge.

 

[cyberjock]

Out of how many? Sounds like someone viewed the "fragile" sticker as a challenge.

LOL!

 

[IanWorthington]

12. Or 15 if you count the replacements (yeah, I'm running w/o a spare right now...)

 

[Ericloewe]

12. Or 15 if you count the replacements (yeah, I'm running w/o a spare right now...)

1/3 DoA rate? That is nasty...

LOL!

A small study some time ago showed that packages labeled "Fragile" actually received worse treatment than "normal" packages. Not the most significant of studies, but still a rather unfortunate conclusion...