Log file rotation too slow

[Stanley Liu]

BIND 8.2.3-REL-NT3

The following is the logging configurations in the named.conf file:

logging
{
 channel other_log {
 file "c:\winnt\system32\dns\etc\other.log" versions unlimited size 20m;
 severity debug;
 print-time yes;
 };
 channel query_log {
 file "c:\winnt\system32\dns\etc\query.log" versions unlimited size 20m;
 severity debug;
 print-time yes;
 };
 category default { other_log; };
 category lame-servers { null; };
 category cname { null; };
 category update { other_log; };
 category security { other_log; };
 category panic { other_log; };
 category packet { other_log; };
 category eventlib { other_log; };
 category queries {query_log; };
};

Every time the log file is full, it "renames" the oldest log file from, say
query.log.30 to query.log.31; then query.log.29 to query.log.30 and so on until
it "renames" the current log file query.log to query.log.0 and starts a new log
file.  During this period, Named does not response to query at all.  To add to
that, it takes about 10 seconds to "rename" one log file.  So if I have 100 log
files (maximum), it will take around 20 minutes to "rename" the whole lot which
means that it will not response to query for 20 minutes.  Why it takes 10
seconds to "rename" a file? or is it a file "copy"?  or is it a NT port problem?
or is it just me?  How do you people handle logging and the versioning of it?

--

Stanley Liu
stanley.liu at tmca.com.au