secondary lag after NOTIFY

[Mathias Körber]

> We use BIND version 8.2.2-P7 on primary and secondary.
> A cron job on the performs an ndc reload at 4am daily.
> We add a lot of zones and at 5am daily named.conf on the
> primary is FTPed over to the secondary.  Some "kung fu"
> then converts the master entries in named.conf into
> slave syntax. An ndc reload then takes place on the secondary.
>=20
> Following a record and SOA change in a primary zone file,
> it can take up to a week for it to appear in the secondary
> copy, in spite of logs on both servers showing the primary
> sending the NOTIFY promptly and the secondary ACKing it
> within about 30 mins.
>=20
> Messages of the following format appear in the secondary log,
> however inspections of the contents of (in this case)
> /home/named/slave/fred.com days later show that no update has
> taken place:
>=20
> "notify: NOTIFY(SOA) for zone already xferring (fred.com)"

While this does not directly address this message, you may want to start =
by
specifying a serial-queries option with a fairly large argument (say 500 =
for
starters, maybe up to the low thousands) in the slave's named.conf. This
will allow the slave to process this many serial# queries at the same =
time,
which it does after notifies. If this is not increased from the default,
it will only process a low (can't remember the exact number right now,
check the source) default, which may delay detection of changed zones =
even
with notify on if the slave has many zones to handle.