bind 8.2.3 memory leak?
Kevin Darcy
kcd at daimlerchrysler.com
Thu Apr 19 19:53:07 UTC 2001
First of all, what's "8.2.3-NOISE=8+fix1152"? Is that some SunSolve-patched
version of BIND 8.2.3? If you're not running a vanilla 8.2.3, then who knows if
there's some proprietary code that might have some sort of memory leak in it?
Alternatively, why don't you try dumping your cache periodically? Maybe you
overlooked something in your config and perhaps you *are* in fact caching entries
after all. Maybe your clients are hitting the server with a bunch of Internet
names and your memory is filling up with negative cache entries? Dumping the cache
will reveal all.
- Kevin
jackszewczyk at aol.com wrote:
> Hi,
>
> Environment: Solaris 2.5/bind 8.2.3/sparc
>
> =========================================
> [dns-eccc-a]/opt/local/etc/namedb/special{251}(root)#what /local/sbin/named
> /local/sbin/named:
> Copyright (c) 1986, 1989, 1990 The Regents of the University of California.
> named 8.2.3-NOISE=8+fix1152 Tue Feb 27 15:06:36 EDT 2001
> root at dnsmaster:/com/pub/bind/bind823/bin/named
> SunOS 5.5 Generic November 1995
> =========================================
>
> Problem Description:
>
> Client has 6 DNS servers. named on this servers can run for a couple of days
> and then crashes. None of these servers is a caching server (none of them
> actually
> has got any "hint" directive in configuration files - they are all authoratives
> for internal DNS domain and don't know anything about external domaines. At the
> moment
> they restart named every 24 h to workaround memory/crash issue.
>
> I wrote a simple script to monitor named memory usage:
>
> Script:
> =========================================================
> #!/bin/sh
> ##########################################################
> # name: test_mem_named.scr
> # description: script to monitor memory usage by named 8.2.3
> # run by cron every 15 min.
> # etime: total elapsed time since process was started
> # vsz: total size in virtual memory in [kb]
> # osz: size of swappable memory of the process [pages]
> # rss: total size in resident memory in [kb]
> # pmem: resident portion of the process/total physical
> # memory of the system in [%]
> # author: JS 18-04-2001
> ##########################################################
> cd /local/users/dnsadmin/test
> ps -e -o etime,comm,vsz,osz,rss,pmem | grep named >> named.mem
> =========================================================
>
> Output from the script:
> =======================
> etime comm vsz osz rss pmem
> =========================================================
> 14:19:49 /usr/local/sbin/named 25896 6474 12552 20.0
> 14:34:48 /usr/local/sbin/named 26216 6554 12620 20.1
> 14:49:50 /usr/local/sbin/named 26728 6682 12728 20.3
> 15:04:48 /usr/local/sbin/named 26984 6746 12776 20.4
> 15:19:48 /usr/local/sbin/named 27440 6860 12896 20.6
> 15:34:48 /usr/local/sbin/named 27760 6940 12996 20.7
> 15:49:49 /usr/local/sbin/named 28528 7132 13192 21.1
> 16:04:48 /usr/local/sbin/named 29232 7308 13364 21.3
> 16:19:49 /usr/local/sbin/named 29488 7372 13492 21.5
> 16:34:49 /usr/local/sbin/named 30128 7532 3480 5.6
> 16:49:49 /usr/local/sbin/named 30512 7628 4368 7.0
> 17:04:48 /usr/local/sbin/named 30640 7660 5368 8.6
> 17:19:48 /usr/local/sbin/named 31024 7756 5532 8.8
> 17:34:48 /usr/local/sbin/named 31792 7948 5956 9.5
> 17:49:50 /usr/local/sbin/named 32368 8092 6124 9.8
> 18:04:49 /usr/local/sbin/named 32688 8172 6264 10.0
> 18:19:48 /usr/local/sbin/named 33016 8254 6420 10.3
> 18:34:48 /usr/local/sbin/named 33272 8318 6480 10.4
> 18:49:49 /usr/local/sbin/named 33720 8430 6588 10.5
> 19:04:48 /usr/local/sbin/named 33976 8494 6676 10.7
> 19:19:49 /usr/local/sbin/named 34360 8590 6872 11.0
> 19:34:48 /usr/local/sbin/named 34744 8686 6964 11.1
> 19:49:49 /usr/local/sbin/named 35448 8862 7120 11.4
> 20:04:48 /usr/local/sbin/named 35576 8894 7200 11.5
> 20:19:49 /usr/local/sbin/named 35896 8974 7320 11.7
> 20:34:48 /usr/local/sbin/named 36152 9038 7432 11.9
> 20:49:49 /usr/local/sbin/named 36536 9134 7548 12.1
> 21:04:48 /usr/local/sbin/named 36920 9230 7636 12.2
> 21:19:49 /usr/local/sbin/named 38016 9504 8112 13.0
> 21:34:48 /usr/local/sbin/named 38592 9648 8240 13.2
> 21:49:50 /usr/local/sbin/named 38912 9728 8352 13.3
> 22:04:48 /usr/local/sbin/named 39424 9856 8512 13.6
> 22:19:48 /usr/local/sbin/named 39680 9920 8612 13.8
> 22:34:48 /usr/local/sbin/named 40128 10032 8708 13.9
> 22:49:49 /usr/local/sbin/named 40512 10128 8872 14.2
> 23:04:48 /usr/local/sbin/named 40768 10192 8948 14.3
> 23:19:49 /usr/local/sbin/named 41024 10256 9068 14.5
> 23:34:48 /usr/local/sbin/named 41280 10320 9192 14.7
> 23:49:49 /usr/local/sbin/named 41600 10400 9272 14.8
> 4:44 /usr/local/sbin/named 5992 1498 2848 4.6
> 19:45 /usr/local/sbin/named 6504 1626 3188 5.1
> 34:44 /usr/local/sbin/named 6824 1706 4164 6.7
> 49:47 /usr/local/sbin/named 6952 1738 4348 7.0
> 01:04:45 /usr/local/sbin/named 7464 1866 5448 8.7
> 01:19:45 /usr/local/sbin/named 7848 1962 5580 8.9
> 01:34:45 /usr/local/sbin/named 7976 1994 5660 9.1
> 01:49:46 /usr/local/sbin/named 8424 2106 5756 9.2
> 02:04:44 /usr/local/sbin/named 8944 2236 5868 9.4
> 02:19:45 /usr/local/sbin/named 9072 2268 5976 9.6
> 02:34:44 /usr/local/sbin/named 9280 2320 6056 9.7
> 02:49:45 /usr/local/sbin/named 9856 2464 6200 9.9
> 03:04:44 /usr/local/sbin/named 10176 2544 6352 10.2
> 03:19:44 /usr/local/sbin/named 10560 2640 6496 10.4
> 03:34:44 /usr/local/sbin/named 10752 2688 6568 10.5
> 03:49:46 /usr/local/sbin/named 11072 2768 6732 10.8
> 04:04:45 /usr/local/sbin/named 11328 2832 6800 10.9
> 04:19:44 /usr/local/sbin/named 11648 2912 6876 11.0
> 04:34:44 /usr/local/sbin/named 12224 3056 7224 11.5
> 04:49:45 /usr/local/sbin/named 12736 3184 7364 11.8
> 05:04:44 /usr/local/sbin/named 12864 3216 7468 11.9
> 05:19:45 /usr/local/sbin/named 13184 3296 7512 12.0
> 05:34:44 /usr/local/sbin/named 13576 3394 7588 12.1
> 05:49:45 /usr/local/sbin/named 13960 3490 7704 12.3
> 06:04:44 /usr/local/sbin/named 14344 3586 7864 12.6
> 06:19:45 /usr/local/sbin/named 14984 3746 7984 12.8
> 06:34:44 /usr/local/sbin/named 15368 3842 8068 12.9
> 06:49:46 /usr/local/sbin/named 15624 3906 8204 13.1
> 07:04:44 /usr/local/sbin/named 15944 3986 8332 13.3
> 07:19:46 /usr/local/sbin/named 16200 4050 8404 13.4
> 07:34:44 /usr/local/sbin/named 16200 4050 8464 13.5
> 07:49:46 /usr/local/sbin/named 16776 4194 8596 13.7
> 08:04:45 /usr/local/sbin/named 17352 4338 8768 14.0
>
> Please notice that vsz memory allocation grows pretty quickly. I could explain
> this
> if this server was acting as a caching server but it doesn't. Why is rss memory
> allocation growing? This is the memory resident portion of the process.
> Has anyone got a similar problem or maybe knows what's triggering it?
>
> Jack
More information about the bind-users
mailing list