I've got a (new) 8TB SATA hard disk that I bought to backup my Ubuntu 16.04 data disks. It's connected to my box via a Technet USB3 docking station (that I've used before and not had problems with).
My backup script is essentially
/sbin/mkfs.ext4 -c /dev/sdg1
mount /dev/sdg1 /mnt/backup
mkdir -p /mnt/backup/data1
cp -a /my-pool/my-data1/* /mnt/backup/data1
mkdir -p /mnt/backup/data2
cp -a /my-pool/my-data2/* /mnt/backup/data2
umount /dev/sdg1
This time I ran mkfs separately after partitioning the disk rather than as part of the script. mkfs said it was using 4K blocks when formatting the filesystem. It seemed to finish quite quickly compared to when I've formatted smaller drives with ext3
This copies about 4.5Tb from the ZFS filesystems data1 & data2. No errors were shown at all during this process.
Since this was a new disk I thought I would be paranoid and run /sbin/fsck.ext4 against /dev/sdg1 expecting everything would be fine. However, fsck complained that the disk was not cleanly unmounted (even though it was and the system had been powered on continuously) and then proceeded to find large numbers of errors. I got fed up manually responding so I quit fsck and re-ran as /sbin/fsck.ext4 -y /dev/sdg1 and fsck found hundreds of errors (multiply claimed blocks, checksum and others which I did not manage to note down before fsck filled the terminal history).
Does this mean the hardrive itself is likely bad or is there another explanation for this?
My understanding is that umount should have forced any pending blocks to have been written so I don't understand why fsck would claim the drive was not cleanly unmounted.
Are there some special gotchas with ext4 and large drives or something else I'm likely doing wrong?
File ??? (inode #85393418, mod time Thu Jul 20 08:01:35 2017)
has 1 multiply-claimed block(s), shared with 1 file(s):
... (inode #208650243, mod time Tue Dec 24 21:15:08 2097)
Clone multiply-claimed blocks? yes
File ??? (inode #85393421, mod time Thu Jul 20 07:58:56 2017)
has 1 multiply-claimed block(s), shared with 1 file(s):
... (inode #208650243, mod time Tue Dec 24 21:15:08 2097)
Clone multiply-claimed blocks? yes
File ??? (inode #85393422, mod time Thu Jul 20 07:59:38 2017)
has 3 multiply-claimed block(s), shared with 2 file(s):
... (inode #208650243, mod time Tue Dec 24 21:15:08 2097)
... (inode #211472387, mod time Sat Jul 15 23:06:03 2056)
Clone multiply-claimed blocks? yes
File ??? (inode #85393423, mod time Thu Jul 20 07:58:28 2017)
has 3 multiply-claimed block(s), shared with 1 file(s):
... (inode #208650243, mod time Tue Dec 24 21:15:08 2097)
Clone multiply-claimed blocks? yes
Also weird is that the same inode (208650243) appears in all the errors that I've looked at.
UPDATE: fsck.ext4 is still running! I left it going since I felt there was at least a chance it would complete and leave me with a useful filesystem, but it's been going 16+ hours now.
UPDATE2: I had thought it might be because I started using the drive before jbd2 had finished creating all the journalling structure (which apparently happens lazily, see https://askubuntu.com/questions/119742/ext4-jbd2-journaling-active-even-on-empty-filesystem). I wiped the existing partition table, created a new partition and waited for ~6 hours until jbd2 appeared to stpo writing to the disk. I then used rsync to backup the filesystem. All seemed good. I waited an hour or so and then used umount to umount the device. However same error when I tried to re-mount it:
mount: mount /dev/sdg1 on /mnt/bdavepool failed: Structure needs cleaning