Re: musing about performance of various methods of copying (local) filesystems....

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Mon, 23 Dec 2024 13:02:33 -0800
    From:        "Greg A. Woods" <woods%planix.ca@localhost>
    Message-ID:  <m1tPpZR-00Mv9LC@more.local>

  | Would mounting the target filesystem with '-o async' have helped?

Yes.   I do that all the time - if the target filesystem can just be wiped,
and you can start again, with no loss except for some time, in the unlikely
event of some kind of crash during the operation,  then async mounts are
perfect.   The final sync when it is all done can take a while, but at least
the directory and inodes (or log entries referring to them) aren't constantly
being written and rewritten (the filesystem is back like a v7 filesys, more
or less).

  | I wonder too if running multiple processes (e.g. pax -rw's), especially
  | on the writing side, might help

It is possible, but unlikely, you're more likely to just increase contention
for the output device.   If you were writing to multiple different filesystems,
particularly if the output is on a seekless device (ssd, or if anyone
remembers them, a drum).

  | filesystem in the LV inaccessible from the dom0 because the device
  | mapper does not fully emulate a device and thus does not see disk labels
  | within the LV and does/can not make wedges in the dom0

You could try laying a vnd on top of the volume in that case, that should
allow labels (incl gpt) to be seen.

kre