problem with 9.9.6 and file descriptor limits

[PORTER, BLAIR]

Hello, I recently compiled Bind 9.9.6, on RedHatas follows:

BIND 9.9.6-S2 (Subscription Edition) <id:eb5b129c> built by make with '--enable-threads' '--enable-fixed-rrset' '--disable-openssl-version-check' '--with-openssl=no'
compiled by GCC 4.1.2 20080704 (Red Hat 4.1.2-52)

All seemed ok, except for 1 particular DNS server which had problems, getting messages in /var/log/messages at startup:

Dec  1 10:23:24 clpi263 named[1297]: adjusted limit on open files from 16000 to 1048576
Dec  1 10:23:24 clpi263 named[1297]: found 12 CPUs, using 12 worker threads
Dec  1 10:23:24 clpi263 named[1297]: using 6 UDP listeners per interface
Dec  1 10:23:24 clpi263 named[1297]: using up to 4096 sockets
...
...
...
Dec  1 10:23:25 clpi263 named[1297]: socket: file descriptor exceeds limit (4096/4096)
Dec  1 10:23:25 clpi263 named[1297]: set up managed keys zone for view V049, file '/etc/namedb/Data/MKeys/98296d0c75d7c474f605af1e9e8f6bb6c7aa336f12e022a1819a891c73ae34d9.mkeys'

Several of my views on that server had this message.  I also found this in my normal DNS log file:

01-Dec-2014 11:45:35.473 general: critical: adb.c:2926: REQUIRE((options & 0x00000003) != 0) failed, back trace
01-Dec-2014 11:45:35.473 general: critical: #0 0x413e0b in assertion_failed()+0x4b
01-Dec-2014 11:45:35.473 general: critical: #1 0x5c319a in isc_assertion_failed()+0xa
01-Dec-2014 11:45:35.473 general: critical: #2 0x476ed4 in dns_adb_createfind2()+0x1044
01-Dec-2014 11:45:35.473 general: critical: #3 0x5336c7 in findname()+0xe7
01-Dec-2014 11:45:35.473 general: critical: #4 0x5392d5 in fctx_getaddresses()+0x3b5
01-Dec-2014 11:45:35.473 general: critical: #5 0x53ce5a in fctx_try()+0xcca
01-Dec-2014 11:45:35.473 general: critical: #6 0x543bb8 in fctx_start()+0x1a8
01-Dec-2014 11:45:35.473 general: critical: #7 0x5e2d1c in run()+0x2bc
01-Dec-2014 11:45:35.473 general: critical: #8 0x3a91407851 in _fini()+0x3a90e0e889
01-Dec-2014 11:45:35.473 general: critical: #9 0x3a90ce811d in _fini()+0x3a906ef155

File descriptor limit info:

ulimit -nS            1024

ulimit -nH           4096

my named process /proc//limits data is:
Limit                     Soft Limit           Hard Limit           Units
Max cpu time              unlimited            unlimited            seconds
Max file size             unlimited            unlimited            bytes
Max data size             unlimited            unlimited            bytes
Max stack size            unlimited            unlimited            bytes
Max core file size        unlimited            unlimited            bytes
Max resident set          unlimited            unlimited            bytes
Max processes             256388               256388               processes
Max open files            1048576              1048576              files
Max locked memory         65536                65536                bytes
Max address space         unlimited            unlimited            bytes
Max file locks            unlimited            unlimited            locks
Max pending signals       256388               256388               signals
Max msgqueue size         819200               819200               bytes
Max nice priority         0                    0
Max realtime priority     0                    0
Max realtime timeout      unlimited            unlimited            us

I don't get this problem anywhere else, only on this 1 particular DNS server.  It has many different views, and 15x virtual IPs, so I know it is a complex configuration.  My previous version was bind 9.8.6-P1 which ran just fine.  I have read much information, but can't seem to put my finger on the solution.  I strongly suspect some type of O/S limit issue, not a problem with the Bind code.

Help

Blair Porter
AT&T




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/bind-users/attachments/20141202/63613d35/attachment.html>