Tag Archives: HDD

use zdb to remove pesky device from zfs pool

I had the following problem with a device in my pool:


        NAME                                            STATE     READ WRITE CKSUM
        storage                                         DEGRADED     0     0     0
          mirror-0                                      ONLINE       0     0     0
            WORKING_DISK_1  ONLINE       0     0     0
            WORKING_DISK_2    ONLINE       0   0     0
          mirror-1                                      DEGRADED     0     0     0
            WORKING_DISK_3  ONLINE       0     0     0
            replacing-1                                 DEGRADED     0     0     0
              PROBLEM_DISK  FAULTED      6     1     0  too many errors

However when I tried to replace the drive I got this message:

no such device in pool

I found here that you can use zdb to obtain the GUID of the problem device and replace it that way:

root@nas:~# zdb -l PROBLEM_DISK
failed to unpack label 0
    version: 5000
    name: 'storage'
    state: 0
    txg: 5675107
    pool_guid: 8785893899843624400
    errata: 0
    hostname: 'nas'
    top_guid: 9425730683443378041
    guid: 3449631978925631053
    vdev_children: 2
        type: 'mirror'
        id: 1
        guid: 9425730683443378041
        metaslab_array: 41
        metaslab_shift: 35
        ashift: 12
        asize: 4000782221312
        is_log: 0
        create_txg: 4
            type: 'disk'
            id: 0
            guid: 17168510556101954329
            path: 'WORKING_DISK_3'
            devid: 'WORKING_DISK_3_ID'
            phys_path: 'pci-0000:00:1f.2-ata-2'
            whole_disk: 1
            DTL: 14700
            create_txg: 4
            type: 'disk'
            id: 1
    ----->  guid: 3449631978925631053
            path: 'PROBLEM_DISK'
            devid: 'PROBLEM_DISK_ID'
            phys_path: 'pci-0000:00:1f.2-ata-4'
            whole_disk: 1
            DTL: 14699
            create_txg: 4
    labels = 1 2 3 

I used the guid of the problem disk, and all was well:

zpool replace storage 3449631978925631053 NEW_WORKING_DISK

worked instead of complaining the device I was trying to replace didn’t exist.

Check hard drives for bad sectors in Linux/BSD

It turns out that when hard drives fail, they don’t all fail completely. In fact, most fail silently, getting worse and worse as time moves on, causing bitrot and other issues.

I had a suspicion that one of my drives was failing so I thought I would test it. The tool for the job: badblocks.

badblocks writes data to the drive and then reads it back to ensure it gets the expected result. I have learned a lot about hard drive failure lately and now subscribe to running badblocks on every new hard drive I receive to ensure it is a good drive. The command I use is:

badblocks -wsv <device>

This is a destructive write test – it will wipe the disk. You can also run a non-destructive test, but for new disks you can go ahead and wipe them. I also use badblocks to ensure old disks can still be trusted with data. It’s great for “burn in” testing to ensure a drive won’t fail.

Update 3/1/19: If you encounter the following error:

badblocks: Value too large for defined data type invalid end block (5860522584): must be 32-bit value

It means your drive is too big for badblocks to recognize using the default sector size. Fix this by specifying a 4k sector size:

badblocks -b 4096 -wsv <device>

Thanks to Ubuntu Forums for the info.