Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 10 Jun 2002 09:55:26 -0700 (PDT)
From:      Matthew Dillon <dillon@apollo.backplane.com>
To:        "Robert Blayzor" <rblayzor@inoc.net>
Cc:        "Valery G. Utkin" <uvg@niiefa.spb.su>, freebsd-stable@FreeBSD.ORG
Subject:   Re: FW: RE: Swap_pager error
Message-ID:  <200206101655.g5AGtQsE008876@apollo.backplane.com>
References:   <003f01c21079$21013650$6f00000a@z0.inoc.net>

next in thread | previous in thread | raw e-mail | index | archive | help
    Interesting.  So what you are saying is that the crash may be
    related to a cron job running 'ipfw' on a system whos kernel
    does not have IPFIREWALL configured.

    I guess it is possible, though somewhat unlikely.  Another 
    possibility is that with ipfw disabled the error output from 'ipfw'
    being piped to diff, perl, etc are causing one of those programs
    to get confused and do something bad. 

    A third possibility is that the kernel and ipfw binaries for
    the people effected are out of sync.  Running an old ipfw binary
    on a new kernel has historically created trouble (though not a crash
    per-say).

    It should be noted that /etc/security does not exist in 4.6.  The
    security scripts have all been moved to /etc/periodic/security.  This
    isn't to say that the problem might not still exist.  What we need 
    is a definitive test.  My bet is on binaries which are out of sync
    from the kernel.

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>


:Matt,
:
:There is some merit to what Valery has stated below.  We commented out
:/etc/security from running and since then our server has not crashed at
:all.  Perhaps a place to start looking?
:
:Our server does NOT have IPFW enabled.  
:
:--
:Robert Blayzor, BOFH
:INOC, LLC
:rblayzor@inoc.net
:
:Memory dump:  Amnesia...
:
:
:
:> -----Original Message-----
:> From: Valery G. Utkin [mailto:uvg@niiefa.spb.su] 
:> Sent: Thursday, June 06, 2002 4:57 AM
:> To: Robert Blayzor
:> Cc: freebsd-stable@FreeBSD.ORG
:> Subject: Re: RE: Swap_pager error
:> 
:> 
:> I had the same problem after installation FreeBsd 4.5 without 
:> ipfw. After
:> removal from /etc/security lines which start  ipfw and ip6fw, 
:> all works
:> normally.
:> Valery G. Utkin
:> 
:> ----- Original Message -----
:> From: "Robert Blayzor" <rblayzor@inoc.net>
:> To: "'Matthew Dillon'" <dillon@apollo.backplane.com>;
:> <freebsd-stable@FreeBSD.ORG>
:> Sent: Wednesday, June 05, 2002 5:07 AM
:> Subject: RE: RE: Swap_pager error
:> 
:> 
:> > I looked through all the periodic daily stuff.  It doesn't 
:> seem that any
:> > of the scripts will trash NFS mounted partitions, almost 
:> everything I
:> > saw would only look at UFS mounted partitions.
:> >
:> > One thing I did notice is the security check was quite 
:> brutal.  While
:> > any server should survive it I believe this is what is causing the
:> > system to crash.  The security check seems to run a find on the NFS
:> > servers local UFS mounts.  We have some very, very large 
:> volumes with
:> > hundreds of thousands of small files... (maildirs, boxes, 
:> webmail, web,
:> > etc).  On this box, it seemed that the security check would 
:> take almost
:> > 3-4 minutes to complete with that find, and it just totally 
:> saturates
:> > the box in activity when it runs.
:> >
:> > So, I think there may be a loading issue with all these 
:> files/inodes in
:> > relation to the find process... Perhaps the SCSI or driver stuff in
:> > FreeBSD.  If I can be of any help on this, I surely will 
:> led a hand.  I
:> > would like to see FreeBSD be able to survive this without a hitch.
:> >
:> > Perhaps a suggestion to change the priority of the "find" 
:> tasks in those
:> > scripts with nice or something.  I mean the box was really 
:> bogged down
:> > when we ran "periodic daily" by manually.
:> >
:> > As a work around, we moved periodic daily to run at 9:01am 
:> instead of
:> > 3:01am, and only on Monday - Friday.  We don't need any more weekend
:> > surprise pages and then call-ins.  :-)
:> >
:> > Since this box is an internal server only with no accounts 
:> on it, and it
:> > has no route to the outside + behind a firewall, we're 
:> going to go ahead
:> > and disable the security check all together.  I'm hoping 
:> that this will
:> > provide a work around for this "loading" issue.  If I can 
:> be any help to
:> > the core team to debug this problem, I'll do my best to do 
:> what I can.
:> >
:> > --
:> > Robert Blayzor, BOFH
:> > INOC, LLC
:> > rblayzor@inoc.net
:> >
:> > One picture is worth 128K words.
:> >
:> >
:> >
:> > > -----Original Message-----
:> > > From: Matthew Dillon [mailto:dillon@apollo.backplane.com]
:> > > Sent: Tuesday, June 04, 2002 1:49 PM
:> > > To: Robert Blayzor; freebsd-stable@FreeBSD.ORG
:> > > Subject: Re: RE: Swap_pager error
:> > >
:> > >
:> > >     I have one more idea... daily cron jobs tend to really
:> > > load down the
:> > >     system for a short period of time, especially the disks.
:> > > In your case
:> > >     the local daily cron is combinging with the daily cron
:> > > running on the
:> > >     NFS clients.  There could be a hardware problem with the
:> > > system that
:> > >     is most likely to show up under heavy loads.
:> > >
:> > >     It is also possible that this is revealing a driver 
:> bug somewhere.
:> > >     For example, the extreme disk load could be revealing 
:> a bug in the
:> > >     driver's tag handling or in the RAID card's tag handling.
:> > >  The lack
:> > >     of driver-based error messages is rather odd.  I don't
:> > > see how that
:> > >     can happen unless the RAID card itself is locking up.
:> > >
:> > > -Matt
:> > >
:> > > ::
:> > > ::Both times the box has crashed crashed at ~3:02am.  I'm
:> > > thinking that
:> > > ::something in periodic daily is causing the crashes.
:> > > ::
:> > > ::Keep in mind, that this server serves several NFS clients
:> > > which mount
:> > > ::things such as FreeBSD ports and /usr/src.  Those are soft
:> > > linked to on
:> > >
:> >
:> >
:> > To Unsubscribe: send mail to majordomo@FreeBSD.org
:> > with "unsubscribe freebsd-stable" in the body of the message

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200206101655.g5AGtQsE008876>