-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
And speaking of data migrations...
I performed an ndmpcopy of a filesystem to move it off of a set of DS14 Mk1 shelves (72 GB FC drives, which are now end of support) to a newer set of DS14Mk4 shelves containing 300 GB FC drives. The problem is that after ndmpcopy has finished, the source and destination volume sizes are different:
filer*> df -k old_vol Filesystem total used avail capacity Mounted on /vol/old_vol/ 547880960KB 401694484KB 146186476KB 73% /vol/old_vol/
filer*> df -k new_vol Filesystem total used avail capacity Mounted on /vol/new_vol/ 547880960KB 401639832KB 146241128KB 73% /vol/new_vol/
So we're looking at ~50 MB of difference between the two. After doing filesystem checks on both, I found that both contain the same number of files, all of which are equivalent in size. The directory structures are also equivalent, with one exception: one directory is 12KB on the source and 4KB on the destination. Another anomaly is that a 'df' of the filesystems on Solaris 10 show a difference of 16 files between the source and destination.
/usr/SAG/db-backup (filer:/vol/new_vol/db-backup):292482200 blocks 19023148 files /mnt/old_vol (filer:/vol/old_vol/db-backup):292372920 blocks 19023132 files
So, I know that ndmpcopy is a logical (not physical) copy of the volume, but how is that accounting for more files with fewer blocks? Is it an artifact of fragmentation and metadata changes in WAFL when the data is dumped? I don't have a good explanation for my customer, so I'm hoping someone here will have a quick and easy answer.
Thanks,
- -=Tom Nail