BIND - in loop rewrite zone serial no.

Tony Finch dot at dotat.at
Thu Jan 30 11:21:57 UTC 2020


Milan Jeskynka Kazatel <KazatelM at seznam.cz> wrote:
>
> could someone, please, help me with diagnostics, how can I check how many
> records are signed per cycle?

I looked at my zone transfer logs, which give the size of each IXFR
following a zone update. If you don't have any ixfr logs, then you can use
`named-journalprint` and look at the number of consecutive "add" and
"del" lines in each update.

The reason the sig-signing-* options don't have much effect on the size of
signing increments is a combination of two things: Firstly, `named` adds
random jitter to re-signing times to spread out the load, and when this is
combined with small sig-signing-* values and incremental updates, the
signing schedule will become increasingly fragmented; Secondly, there
isn't any option to de-fragment signing times, so that small updates get
re-signed in bigger chunks.

I inherited a patch from Chris Thompson to increase the hardcoded value
in `stop = now + 5` (see link below). This is what controls how much
defragmentation can happen - 5 seconds is not much! IIRC we adjusted it to
300 (5 minutes) but even something like 3600 would be reasonable, I think,
because then the sig-signing-* options would be the main limit on
re-signing work.

https://gitlab.isc.org/isc-projects/bind9/blob/master/lib/dns/zone.c#L6787

I no longer use this patch because it's easier to run unpatched BIND in
production, and I didn't think it was important enough to turn into a
configuration option. Maybe I was wrong :-)

Tony.
-- 
f.anthony.n.finch  <dot at dotat.at>  http://dotat.at/
Irish Sea, Shannon: Southwest 5 to 7, increasing gale 8 at times. Moderate or
rough in Irish Sea, rough or very rough in Shannon. Rain at times, fog patches
at first in south. Moderate or good, occasionally very poor.


More information about the bind-users mailing list