[Bug 219049] ftruncate() slow on UFS+journaled softupdates

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Fri Feb 23 03:36:43 UTC 2018


Alexander Motin <mav at FreeBSD.org> changed:

           What    |Removed                     |Added
                 CC|                            |mav at FreeBSD.org

--- Comment #7 from Alexander Motin <mav at FreeBSD.org> ---
Could somebody experiencing the problem to look on the disk which was
originally reported as 100% busy during the test with `gstat -I 1s -pod`?  It
would show distribution of I/O operations between reads, writes, deletes and
cache flushes, providing counts, sizes and average latency for which.  It could
be a starting point is the problem is indeed disk-bound.

As I understand, journaled UFS may started to use cache flushes, not used
before, and I can theoretically guess some scenarios when some disks are not
handling it well.

You are receiving this mail because:
You are the assignee for the bug.

More information about the freebsd-fs mailing list