Bind cache size

Daniel Ryslink daniel.ryslink at
Thu Jun 8 12:48:09 UTC 2006


Bind 9.3.2 fell into a strange state today - it returned answer to every 
DNS query with the ANSWER section of the packet empty.

named.log and client.log state that there was not enough memory for 
resolving more recursive queries:


08-Jun-2006 14:07:48.505 client: warning: client 
2001:1508:1000:7::2#59016: no more TCP clients: out o
f memory

08-Jun-2006 13:36:48.606 general: error: isc_socket_accept() failed: 
unexpected error
08-Jun-2006 13:36:51.299 general: error: timer.c:650: unexpected error:
08-Jun-2006 13:36:51.300 general: error: couldn't allocate event
08-Jun-2006 13:40:58.416 notify: info: client 
received notify for zone ''
08-Jun-2006 13:40:58.416 general: info: zone refused notify 
from non-master:
08-Jun-2006 14:08:39.878 general: warning: dropped command channel from out of memory
08-Jun-2006 14:08:43.787 general: warning: dropped command channel from out of memory


Even rndc ceased to work due to lack of memory. Strange thing is, the 
machine had more than 500MB free in the time of the incident.

How large cache files did you experience? IMO the bind should be allowed 
to take as much memory as needed for caching purposes to achieve optimum 
performance. However, after exceeding 450MB, I always start to experience 
problems - bind delays answers occassionally.

Do you limit cache size or let it grow freely?

Thank you for any advices.

Daniel Ryslink

More information about the bind-users mailing list