only one named-xfer at a time
Cricket Liu
cricket at nxdomain.com
Fri Aug 3 21:45:40 UTC 2001
> We thought so. 95% of our zones are transferring from our own
> primary and its pretty smooth and fast. The rest are from third
> party masters and we noticed that there were cases where a third
> party master went dark, defunct whatever and the default (2 hours)
> would end up with 5 or 10 named-xfers clogging up and doing nothing.
>
> Our reasoning was by setting the timeout low, a stuck xfer would
> be reaped quickly and retried later.
I thought your reasoning might be something like that, after looking at your
email address. I think that BIND 9's max-transfer-idle-in might be closer
to what you're looking for, but I'm sure you have your reasons for sticking
with BIND 8.
> We're working with updwards of 30,000 zones so we're trying to make
> it as efficient as possible.
>
> You think if we set that timeout higher more xfers would end up
> running concurrently?
Well, my thought was that the single named-xfer process running might be
due to interaction between setting max-transfer-time-in to one and BIND
8's mechanism for dithering zone transfers. BIND 8 tries to limit the
number
of transfers it starts per unit time. If each transfer only lasts a minute
at the
most, maybe that would explain what you're seeing.
That's only a guess--I didn't look closely at the code.
> (P.S we just picked up edition 4 of your book from the ORA con, long time
> fans, there's a copy on almost every desk here :)
Thanks!
cricket
More information about the bind-users
mailing list