Odd named behavior upon startup (8.2.4 on FreeBSD)

[Barry Margolin]

In article <9j5rn3$lss at pub3.rc.vix.com>,
Forrest Aldrich  <forrie at forrie.com> wrote:
>I've posted about this previously, but never got a response.

I saw a message from Cricket Liu dated about an hour or two after your
message.

>We have a problem with the latest BIND 8.2.4 (and I recall this happening 
>with earlier versions) where when the server first starts up, named drags 
>very slowly and occupies 98% of CPU time, even after all the zones are loaded.
>
>This is remedied by sending a HUP signal to the process:
>
>su-2.05# top -o size -b | grep named
>   140 root      48   0 38564K 38108K RUN     41:28 93.95% 93.95% named
>su-2.05# killall -HUP named
>su-2.05# top -o size -b | grep named
>   140 root       2   0 38820K 38392K RUN     41:46 88.09% 88.09% named
>su-2.05# top -o size -b | grep named
>   140 root       2   0 38820K 38392K select  41:47 71.58% 71.58% named
>su-2.05# top -o size -b | grep named
>   140 root       2   0 39140K 38652K select  41:48 63.57% 63.57% named
>su-2.05# top -o size -b | grep named
>   140 root       2   0 40804K 40196K select  42:06  5.81%  5.81% named
>
>Now this doesn't make any sense to me.  I'm trying to understand WHY this 
>is happening.  There is no information I see that's unusual in the logs.

After it finishes loading all the zones and reports "Ready to answer
queries", BIND sends out NOTIFY messages to all the slave servers for each
zone that it loaded, and the slaves all reply to these messages.  If you
have lots of zones, this can load it pretty heavily.

A few weeks ago we changed our master server configuration to include
"also-notify" entries for all our stealth slave servers (last year we
replaced a whole bunch of servers inherited from near.net, barrnet.net,
genuity.net, and sura.net with new sys.gtei.net servers, but we also have
stealth servers running at the old addresses for backward compatibility).
So we went from sending NOTIFY messages to 3 servers per domain to about a
dozen servers.  When we noticed this, I noticed that restarting our server
with about 8,000 zones on it would take several hours, and the log was full
of messages about sending NOTIFY and receiving NOTIFY responses.

-- 
Barry Margolin, barmar at genuity.net
Genuity, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.