[prev in list] [next in list] [prev in thread] [next in thread] 

List:       bind-workers
Subject:    9.3.2b1 timer.c:650: unexpected error: couldn't allocate event
From:       Stefan Schmidt <zaphodb--bind () zaphods ! net>
Date:       2005-12-05 12:25:44
Message-ID: 20051205122544.GO32251 () zaphods ! net
[Download RAW message or body]

Making use of ISC_INTERNAL_MALLOC and --disable-threads i experienced the
following _interesting_ behaviour with BIND 9.3.2rc1 on linux:
I had tons of the following errors in the log:
Dec  4 23:37:20 cns1 named[2219]: timer.c:650: unexpected error:
Dec  4 23:37:20 cns1 named[2219]: couldn't allocate event
Dec  4 23:37:20 cns1 named[2219]: client 194.97.55.191#52438: ns_client_replace() \
failed: out of memory of course for many different IPs. Around 90% of all queries \
were answered with SERVFAIL. Restart of the named process did help but as all of my \
monitoring systems seem to have been the lucky winners of this 'who gets an answer'
contest i was not aware of the situation until i read about it in mails from
collegues in the morning. ;) 
The machine runs Debian/unstable and BIND9 exclusively, has 512MB of RAM and
these are the relevant BIND9 configuration settings:
options {
	max-ncache-ttl 120;
	lame-ttl 1800;
	max-cache-size 300M;
	recursive-clients 8000;
	tcp-clients 300;
	datasize 390M;
	cleaning-interval 300;
}
Just around 5 local zones, such as 127.in-addr.arpa and localhost - i.e.
they are small.
I have some acls, including one for blackholing bogon netranges.
No views at all.
cmdline is: /usr/sbin/named -n 1 -t /var/named -u bind
libc6-i686 2.3.5-6
gcc 4.0.2-1
./configure --prefix=/usr --sysconfdir=/etc/namedb --localstatedir=/var \
--disable-threads --enable-libbind --enable-ipv6 --with-openssl=no

best regards,
		Stefan
-- 
In The Beginning there was nothing, which exploded.


[prev in list] [next in list] [prev in thread] [next in thread]