Once an hour the secondary hangs.

[Jim Reid]

>>>>> "Ted" == Ted Watson <t405405 at hotmail.com> writes:

    Ted> I've an 8.2.3 secondary server on FreeBSD 3.2. This server
    Ted> handles about 8 million queries a day. One month ago the
    Ted> server began hanging for a few minutes at the top of the
    Ted> hour. He would not handle any queries and I could see BIND
    Ted> was not responding for about 3 minutes. I tried to determine
    Ted> if this was a network or system performance issue by pinging
    Ted> out to the net and running traceroutes. I could see at a
    Ted> point the traceroutes would not supply names to local routers
    Ted> but only IP numbers. I upgraded the BIND to 8.2.3 and the
    Ted> problem went away. Now one month later I am again having
    Ted> strange hourly timeouts.

This is probably just the name server preening its cache. It does this
every hour by default. While it's freeing the memory used by expired
RRs the server doesn't answer queries. If this overhead is becoming
noticeable, you should think about a better DNS architecture as a
long-term solution to the problem. [eg Have extra name servers to
share/spread the load of queries from your clients.] A short term
measure would be to try a lower cleaning interval. That way the
outages will hopefully be shorter. However they will be more frequent.

Even so this is unusual. How big is your name server? How much RAM/VM is
it using? Is it thrashing? Why is it getting so many queries? Who's
making them? What sort of TTLs are on those names? Could they be
increased? [These are rhetorical questions.]

    Ted> Could this be some kind of zone transfer failure?

No. Firstly, zone transfers don't have any bearing on a name server's
ability to handle queries, whether or not they fail. [Well not unless
you're running a version of BIND from the Stone Age.] Secondly if zone
transfers were failing there would be error reports in the server's
logs. Have you checked them for any error messages?