BIND - out of memory

Wed Mar 25 06:46:55 UTC 2009

Hi,

more findings ...

BIND 9.6.1b1

No matter what I set in named.conf, it starts to give "out of memory" when recursive
clients pass 1000. I see that 1000 is the default value for recursive-clients.

 From "rndc status" on each run, it starts with "out of memory messages" when recursive-clients
passes 1000:
        recursive clients: 1029/149900/150000

 From named.conf:
        clients-per-query 250;
        max-clients-per-query 1500;
        recursive-clients 50000;

BIND does not allocate space to more than 1000/default clients no matter what is specified in 
named.conf?

I tried it against 9.4.3 also (same config), and the same thing happens:

        nsXX(root) named-new 948# tail -f named-new.log  | grep memory
        25-Mar-2009 07:35:00.504 database: adb: fetch of 'ns-kiev.km.ua' A failed: out of memory
        25-Mar-2009 07:35:00.505 database: adb: fetch of 'y.ns.verio.net' A failed: out of memory
        25-Mar-2009 07:35:00.506 database: adb: fetch of 'dns1.gla.ac.uk' A failed: out of memory
        --cut--

        recursive clients: 1002/49900/50000

.. it starts output "out of memory" messages when It passes 1000 recursive clients.

So it definitly seems that 1000/default recursive clients is the magic limit regarding the "out of memory" 
messages I get.

BIND perhaps allocate space for 1000 recursive-clients, but does not allocate more when named.conf
has another number for it?

Or?

Regards
Jan Arild Lindstrom

At 11:44 24/03/2009, Jan Arild Lindstrøm wrote:
>Hi,
>
>I am running ResPerf from Nominum against BIND 9.6.1b1, and I get a lot of:
>
>--cut--
>24-Mar-2009 08:51:30.495 database: adb: fetch of 'ns2.state.oh.us' A failed: out of memory
>24-Mar-2009 08:51:30.630 database: adb: fetch of 'gz-dns.cncnet.net' A failed: out of memory
>24-Mar-2009 08:51:30.657 query-errors: fetch completed at resolver.c:2908 for 129.83.61.195.in-addr.arpa/PTR in 22.401385: out of memory/success [domain:61.195.in-addr.arpa,referral:2,restart:1,qrysent:0,timeout:0,lame:0,neterr:0,badresp:0,adberr:2,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.672 query-errors: fetch completed at resolver.c:2908 for 211.121.239.211.in-addr.arpa/PTR in 18.586241: out of memory/success [domain:239.211.in-addr.arpa,referral:2,restart:1,qrysent:1,timeout:1,lame:0,neterr:0,badresp:0,adberr:2,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.684 database: adb: fetch of 'iit.rit.ac.th' A failed: out of memory
>24-Mar-2009 08:51:30.685 database: adb: fetch of 'ritk6.rit.ac.th' A failed: out of memory
>24-Mar-2009 08:51:30.708 query-errors: fetch completed at resolver.c:2908 for 118.95.219.66.in-addr.arpa/PTR in 31.293651: out of memory/success [domain:95.219.66.in-addr.arpa,referral:1,restart:3,qrysent:0,timeout:1,lame:0,neterr:0,badresp:0,adberr:0,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.714 query-errors: fetch completed at resolver.c:2908 for 30.126.138.63.in-addr.arpa/PTR in 28.681399: out of memory/success [domain:138.63.in-addr.arpa,referral:1,restart:3,qrysent:0,timeout:1,lame:0,neterr:0,badresp:0,adberr:6,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.715 query-errors: fetch completed at resolver.c:2908 for 161.112.185.194.in-addr.arpa/PTR in 18.591808: out of memory/success [domain:185.194.in-addr.arpa,referral:1,restart:1,qrysent:1,timeout:1,lame:0,neterr:0,badresp:0,adberr:0,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.739 query-errors: fetch completed at resolver.c:2908 for ppp85-141-184-239.pppoe.mtu-net.ru/A in 14.649606: out of memory/success [domain:mtu-net.ru,referral:1,restart:1,qrysent:0,timeout:0,lame:0,neterr:0,badresp:0,adberr:1,findfail:0,valfail:0]
>24-Mar-2009 08:51:30.812 database: adb: fetch of 'tirant.gva.es' A failed: out of memory
>24-Mar-2009 08:51:30.814 database: adb: fetch of 'ns1.pldi.net' A failed: out of memory
>24-Mar-2009 08:51:30.898 database: adb: fetch of 'ns1.corporatecolo.com' A failed: out of memory
>24-Mar-2009 08:51:30.899 database: adb: fetch of 'ns1.gratisdns.dk' A failed: out of memory
>--cut--
>
>What does "database: adb: .. out of memory" mean?
>What does "query-errors: fetch completed at ... out of memory/success" mean?
>
>Solaris 10 on a Sun T5140 with 6 cores/96 threads and 16GB of memory: 
>        SunOS xxx.xxx.xx 5.10 Generic_138888-01 sun4v sparc SUNW,T5140 Solaris
>
>The named process takes only 170MB:
>        Memory: 16G phys mem, 11G free mem, 4104M total swap, 4104M free swap
>        19563 named     99  59    0  171M  169M sleep    1:35  0.00% named
>
>BIND 9.4.3 on the same server (running at the same time as testing 9.6.1b1):
>        10186 named     99  54    0 2990M 2989M cpu/66 5438.0  3.84% named
>
>I tried:
>        datasize unlimited;
>        stacksize unlimited;
>        max-cache-size unlimited;
>
>But it had no effect, I still get just as many "out of memory" lines when running ResPerf.
>        resperf -d queryfile-example-3million -e -s <IP address> -m 10000
>
>Plimit reports (on the named process):
>        resource              current         maximum
>        time(seconds)         unlimited       unlimited
>        file(blocks)          unlimited       unlimited
>        data(kbytes)          unlimited       unlimited
>        stack(kbytes)         unlimited       unlimited
>        coredump(blocks)      unlimited       unlimited
>        nofiles(descriptors)  unlimited       unlimited
>        vmemory(kbytes)       unlimited       unlimited
>
>Any hints on what these "out of memory" messages mean would be appreciated.
>
>Thanks
>Jan Arild Lindstrom
>
>_______________________________________________
>bind-users mailing list
>bind-users at lists.isc.org
>https://lists.isc.org/mailman/listinfo/bind-users