https://bugzilla.novell.com/show_bug.cgi?id=617437
https://bugzilla.novell.com/show_bug.cgi?id=617437#c2
Martin Konold changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |external.martin.konold@de.b
| |osch.com
--- Comment #2 from Martin Konold 2010-12-09 17:07:56 CET ---
I just observed a very similar situation with reiserfs on a sata disk
immediately after I wrote about 1 GB to disk.
Observeration:
rsync -aDogv remote:/source /mnt
umount /mnt <-- takes up to 5 minutes!!
Surprise Workaround:
rsync -aDogv remote:/source /mnt
sync <-- takes about 1 second
umount /mnt <-- takes less than 1 second
One would expect that both scenarios take about the same time while the second
could be slightly slower. Measurement shows that this is not the case.
So doing a sync right before the sycall to umount helps enormously.
Maybe such a sync should be added to umount from util-linux right before
doing the umount syscall as an intermediate workaround in Factory?
IMHO this would not cause much harm while helping many people a lot.
(Not everyone is patient enough to wait up to several minutes for a simple
umount on an otherwise idle system with no pending data to be written to disk.)
I guess that the following stuff
https://bugzilla.kernel.org/attachment.cgi?id=26224 is related to the problem.
Ludwig: Can you please time the following script with your test setup and
report the result?
#!/bin/bash -xe
img="$1"
rm -f "$img"
dd if=/dev/zero of=$img bs=1G count=0 seek=4
mkfs.ext2 -q -F "$img"
mount -o loop "$img" /mnt
dd if=/dev/zero of=/mnt/file bs=1M count=300
time sync
time umount /mnt
Yours,
-- martin
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.