• Hang in umount in poudriere run

    From Kurt Jaeger@pi@freebsd.org to muc.lists.freebsd.stable on Sun Jul 20 10:51:29 2025
    From Newsgroup: muc.lists.freebsd.stable

    Hello,

    I have hanging umount processes in the last two poudriere runs
    on a 14.3p1 server.

    See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=288345

    The umount processes can not be killed:

    20104 2 D+ 0:00.00 umount -f /pou/data/.m/143-default/21
    20156 6- T+ 0:00.00 umount -f /pou/data/.m/143-default/21/.p

    Any ideas what I can do to debug this ?
    --
    pi@FreeBSD.org +49 171 3101372 Now what ?


    --
    Posted automagically by a mail2news gateway at muc.de e.V.
    Please direct questions, flames, donations, etc. to news-admin@muc.de
    --- Synchronet 3.21a-Linux NewsLink 1.2
  • From Kurt Jaeger@pi@freebsd.org to muc.lists.freebsd.stable on Sun Jul 20 12:34:09 2025
    From Newsgroup: muc.lists.freebsd.stable

    Hello,

    I have hanging umount processes in the last two poudriere runs
    on a 14.3p1 server.

    See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=288345

    The umount processes can not be killed:

    20104 2 D+ 0:00.00 umount -f /pou/data/.m/143-default/21
    20156 6- T+ 0:00.00 umount -f /pou/data/.m/143-default/21/.p

    Any ideas what I can do to debug this ?

    I created a ktrace:

    ktrace -f /tmp/ktr3 df
    kdump -f /tmp/ktr3 > df-ktrace

    see:
    https://people.freebsd.org/~pi/logs/hang/df-ktrace

    It looks like the last sign of life are some calls to getfsstat() ?

    23969 df RET close 0
    23969 df CALL munmap(0x2c36300f000,0xcf44)
    23969 df RET munmap 0
    23969 df CALL ioctl(0x1,TIOCGETA,0x2c35fccf394)
    23969 df RET ioctl 0
    23969 df CALL getfsstat(0,0,MNT_NOWAIT)
    23969 df RET getfsstat 68/0x44
    23969 df CALL getfsstat(0xf7dddc4a580,0x4dd40,MNT_NOWAIT)
    23969 df RET getfsstat 67/0x43
    23969 df CALL getfsstat(0xf7dddc4a580,0x4dd40,MNT_WAIT)
    --
    pi@FreeBSD.org +49 171 3101372 Now what ?


    --
    Posted automagically by a mail2news gateway at muc.de e.V.
    Please direct questions, flames, donations, etc. to news-admin@muc.de
    --- Synchronet 3.21a-Linux NewsLink 1.2
  • From Craig Leres@leres@freebsd.org to muc.lists.freebsd.stable on Tue Aug 19 10:53:14 2025
    From Newsgroup: muc.lists.freebsd.stable

    On 7/20/25 01:51, Kurt Jaeger wrote:
    Hello,

    I have hanging umount processes in the last two poudriere runs
    on a 14.3p1 server.

    See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=288345

    The umount processes can not be killed:

    20104 2 D+ 0:00.00 umount -f /pou/data/.m/143-default/21
    20156 6- T+ 0:00.00 umount -f /pou/data/.m/143-default/21/.p

    Any ideas what I can do to debug this ?

    I've been seeing random processes getting stuck in D state since
    14.3-RELEASE (I'm currently on p2) on my poudriere build server. This
    happens about twice a month and only while my daily ~1000 package build
    is in progress. I'll either get nagios alerts from things that have
    wedged or notice that the build hasn't completed and has been running
    more for more than 2X the normal build time (7 hours). Today ntpd (and
    others) got stuck. When this happens the only remedy is to reboot
    (including an ipmi reset unless I'm willing to wait a LONG time). Then
    cleanup the zfs /.m/ filesystems...

    I'm also curious what info I might collect before rebooting.

    Craig


    --
    Posted automagically by a mail2news gateway at muc.de e.V.
    Please direct questions, flames, donations, etc. to news-admin@muc.de
    --- Synchronet 3.21a-Linux NewsLink 1.2
  • From Kurt Jaeger@pi@freebsd.org to muc.lists.freebsd.stable on Tue Aug 19 20:07:22 2025
    From Newsgroup: muc.lists.freebsd.stable

    Hi!

    I have hanging umount processes in the last two poudriere runs
    on a 14.3p1 server.

    See https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=288345
    [...]
    I'm also curious what info I might collect before rebooting.

    See the ticket, I received some hints there.
    --
    pi@FreeBSD.org +49 171 3101372 Now what ?


    --
    Posted automagically by a mail2news gateway at muc.de e.V.
    Please direct questions, flames, donations, etc. to news-admin@muc.de
    --- Synchronet 3.21a-Linux NewsLink 1.2