Home / server / Questions / 1166441
Accepted
slackline
Asked: 2024-10-11 00:42:15 +0800 CST

Replacing RAID1 disk - partition sizes differ

  • 772

One of my RAID1 disks failed recently. I've purchased a replacement, connected it up and formatted it as Linux RAID using cfdisk.

  • New drive is /dev/sdc
  • Old drive is /dev/sdd

They appear to have the same number of bytes and sectors overall but the formatted partitions differ in size...

# fdisk -l /dev/sd{c,d}
Disk /dev/sdc: 7.28 TiB, 8001563222016 bytes, 15628053168 sectors
Disk model: ST8000VN0022-2EL
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes
Disklabel type: gpt
Disk identifier: 952AC4E6-20B9-408D-83CE-AC4AEFBC1075

Device     Start         End     Sectors  Size Type
/dev/sdc1   2048 15628052479 15628050432  7.3T Linux RAID


Disk /dev/sdd: 7.28 TiB, 8001563222016 bytes, 15628053168 sectors
Disk model: ST8000VN004-2M21
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes
Disklabel type: gpt
Disk identifier: 4069B8CB-4C43-304B-B872-DDB89C0E0AA1

Device     Start         End     Sectors  Size Type
/dev/sdd1   2048 15628053134 15628051087  7.3T Linux RAID

Specificly...

/dev/sdc1   2048 15628052479 15628050432  7.3T Linux RAID
/dev/sdd1   2048 15628053134 15628051087  7.3T Linux RAID

The new partition /dev/sdc1 is smaller than the existing partition. As a consequence...

# mdadm --manage /dev/md127 --add /dev/sdc1
mdadm: Value "kimura:0" cannot be set as devname. Reason: Not POSIX compatible. Value ignored.
mdadm: /dev/sdc1 not large enough to join array

I've found a thread suggesting one (/dev/sdd which is the ST8000VN004) is using 512E to emulate 512bytes per logical sector despite being 4096 bytes per physical sector.

Now I'm wary of jumping in with resizing the existing partition on /dev/sdd1 to match that of /dev/sdc1 as this is part of a RAID1 array with three logical volumes on top.

Found a related thread but the issue there is different as they've a 3TB RAID and they want to partition a new 14TB so space isn't an issue, here my new partition is smaller than the existing.

Any thoughts or advice on how to proceed would be appreciated.

raid

1 Answers

  1. Best Answer

    Instead of creating a partition table on the new drive from scratch, copy it from the existing drive using sfdisk:

    sfdisk -d /dev/sdd > part_table
grep -v ^label-id part_table | sed -e 's/, *uuid=[0-9A-F-]*//' | sfdisk /dev/sdc

    This will ensure that partitions are exactly the same.

    Both your disks are the same with regard to the physical sector size. Both are having 4k sectors, according to the output:

    Sector size (logical/physical): 512 bytes / 4096 bytes
I/O size (minimum/optimal): 4096 bytes / 4096 bytes

    While it is indeed not very good that the partition has fractional count of physical sectors, this shouldn't be a problem in practice. It is only important it begins at the correct place. The mdadm and LVM on-disk structures have large paddings as to round boundaries to 64k or even to 4M, so the extra space that falls into incomplete physical sector at the end of the volume simply isn't used.

    However, it might be good to clean this now, for nobody to be confused in the future. If you are using LVM, it's possible to tidy it all without dangerous surgery, using the following sequence:

    1. create degraded RAID1 array out of your new smaller partition
    2. create PV on it
    3. add it into the existing VG
    4. pvmove everything there emptying the old PV,
    5. remove the old "odd" disk from the VG and remove PV labels
    6. repartition it into better scheme and add it back into array.
    • 1