Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 31 Aug 2002 13:03:07 -0700 (PDT)
From:      Matthew Dillon <dillon@apollo.backplane.com>
To:        "Marc G. Fournier" <scrappy@hub.org>
Cc:        Arnvid Karstad <arnvid@karstad.org>, <bmah@FreeBSD.ORG>, <freebsd-stable@FreeBSD.ORG>
Subject:   Re: Problems with FreeBSD - causing zalloc to return 0 ?!
Message-ID:  <200208312003.g7VK37aS002117@apollo.backplane.com>
References:   <20020831140602.O14642-101000@mail1.hub.org>

next in thread | previous in thread | raw e-mail | index | archive | help

:
:  This message is in MIME format.  The first part should be readable text,
:  while the remaining parts are likely unreadable without MIME-aware tools.
:  Send mail to mime@docserver.cac.washington.edu for more info.
:
:--0-1838173444-1030814090=:14642
:Content-Type: TEXT/PLAIN; charset=US-ASCII
:
:
:Morning ... as I said, shouldn't take long :)
:
:I setup the script to log in 1 second intervals, to a log.`date +%S` file,
:hoping that you only wanted the last few seconds of data?  If not, I can
:set it up differently, but here are the result of the last minute of its
:apparent existence ...
:
:First, the kernel panic of:

    Well Marc, there's nothing totally obvious here but it is still quite
    possible that the machine is running out of KVM.

    Try this.  You can run gdb on a running machine as follows:

	gdb -k kernel.debug /dev/mem

    You can then print out elements of the kernel's memory in real time:

    gdb> print kernel_vm_end

    Could you post your kernel config?  And repeat your test but run gdb
    in a shell and print kernel_vm_end at the start of your test and
    every so often while the test is running.

    Other things you can try that will reduce KVM usage:  in
    /boot/loader.conf reduce the amount of physical memory, e.g. reduce
    from 4G to 2G with:

    hw.physmem="2048m"

    and see if you can repeat the problem.  Oddly enough if you cannot
    repeat the problem the issue is almost certainly the kernel running
    out of KVM.  You can also try reducing the number of mbuf clusters
    and reducing maxvnodes, though the values I saw in your logs seem
    reasonable.

    If you only have one swap device you can reducing the number of swap
    devices by setting the NSWAPDEV kernel config variable to 1
    (it defaults to 4).  This will reduce the KVM reservation for 
    the swap bitmap.  You can also reduce the kernel reservation for
    swap block data by setting the kern.maxswzone boot environment
    variable.  This is in bytes, e.g. in /boot/loader.conf

    kern.maxswzone="32m"

    All of these ideas are designed to reduce KVM usage.  I am extremely
    interested in knowing if they have an effect on the crashes you are
    seeing.  If you can stop the crashes from occuring then we are going
    to have to get more conservative in our KVM allocations.

						-Matt


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?200208312003.g7VK37aS002117>