bind-9.6-ESV-R5-P1 crash!!!

Issam Harrathi issamneo at gmail.com
Fri Dec 9 14:58:41 UTC 2011


Hi All,
I discover that bind-9.6-ESV-R5-P1 (configured as a cache-recursor server)
crash in this case:
1/ logging statement are like that:
logging {
        category default { default_syslog; default_debug; };
        category lame-servers { null; };

};


2/ i send 30000 qps using tcpreplay from another server
3/ logrotate is cronned to turn at 4:00 everyday
4/ this configuration generates approximatively 250M by day of data in the
logfile /var/log/messages, And at 4:00 when the logrotate begin and make
the turn of the log (without compress)==> Bind begin a strange behavior
like that:
i made the test 3 times, for the two first i found that bind at 4:00
(before 4:00 all is ok) begin answering only 50% of the querry for 6 hours
aproximatively then bind are no longer running and the server goes in an
OOM (Out Of Memory) du to bind who consume 8 Go of RAM (at 4:00 he consume
just 500 M and begin rising from 500M to 8 Go)!!!!!!!! and the third test
actually running at 4:00 bind begin answering only 50 % of query then at
10:00 he answer all query by a SERVFAIL with a huge query time(2000 msec)

Off Course when i cronned the logrotate to turn each hour the
/var/log/messages doesn't exceed 15M and all is OK. And also when i limit
the data logged (by modifying the logging statement) all is OK too. so why
when the /var/log/messages is about 250 M the bind crash knowing that just
before logrotate begin working I have more than 50 % of RAM, CPU, and
partition Free.



Thanks.
Issam Harrathi.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/bind-users/attachments/20111209/0ff07415/attachment.html>


More information about the bind-users mailing list