Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 2 Oct 2009 10:23:35 -0400
From:      "Grant Peel" <gpeel@thenetnow.com>
To:        <freebsd-questions@freebsd.org>
Subject:   "Out of mbuf address space!"
Message-ID:  <88CD8D18ADF4405DAA81B43F941698A5@GRANT>

next in thread | raw e-mail | index | archive | help
Hi all,

I have an older RAID 5 machine running FreeBSD 5.2.1 and am using it as =
a backup storage unit.

Yesterday morning, we noticed the the NFS mounts on the clients to this =
machine we not available, which sent a bunch of cronjobs spinning out of =
control etc. We also became unable to connect via ssh.

Once at the console we noted sevral dozen entries in the messages.log:

Oct  1 08:32:13 enterprise kernel: Out of mbuf address space!
Oct  1 08:32:13 enterprise kernel: Consider increasing NMBCLUSTERS
Oct  1 08:32:13 enterprise kernel: All mbufs or mbuf clusters exhausted, =
please see tuning(7).

After rebooting the machine, and getting the clients under control I =
started investigating tunning(7) in the man pages.

I am confused however.

I have increased the kern.ipc.nmbclusters to 2048 in the =
/boot/loader.conf, but when I checked netstat -m, it appears that there =
are less buffers available then there were when the problem happened.

netstat -m

enterprise# netstat -m
mbuf usage:
        GEN cache:      0/64 (in use/in pool)
        CPU #0 cache:   145/640 (in use/in pool)
        Total:          145/704 (in use/in pool)
        Mbuf cache high watermark: 512
        Maximum possible: 4096
        Allocated mbuf types:
          144 mbufs allocated to data
          1 mbufs allocated to packet headers
        17% of mbuf map consumed
mbuf cluster usage:
        GEN cache:      0/232 (in use/in pool)
        CPU #0 cache:   135/232 (in use/in pool)
        Total:          135/464 (in use/in pool)
        Cluster cache high watermark: 128
        Maximum possible: 2048                                           =
        <-- this number was much higher
        22% of cluster map consumed                                      =
      <- this number was much lower.
1104 KBytes of wired memory reserved (27% in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

This particular machine has 512 MB of ram.

Any suggestions what an NFS intensive machine with 512 meg ram should =
have kern.ipc.nmbclusters set to?
Are there any otyher tunables I should be looking at.

-Grant




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?88CD8D18ADF4405DAA81B43F941698A5>