Received notify, master unreachable (cached)

[Jarosław Świerczyński]

Hi,

Thank you for reply.

> The whole fact that master sent a notify to slave does NOT mean that the
> master is reachable from the slave, it only means that slave is reachable
> from the master.

In theory you are right. But in most cases it also means that the
master is reachable from the slave. In my opinion BIND should forget
the master was unreachable and simply follow the notification. In the
worst case it will find the master to be unreachable again.

> Either the slave tries to reach the master using wrong source IP address
> (0.0.0.0#0 means it leaves this to system) or the master does not allow
> traffic from slave (e.g. firewall).

Nope, I know for sure that the master becomes reachable from the slave
because I start it. And some time later the slave itself attempts to
contact the master again and it succeeds.

I had a look at the source code and I believe I found the problem. In
lib/dns/zone.c the is a cache where unreachable masters are stored.
Function dns_zonemgr_unreachable() is used to check if a master in
marked as unreachable, while dns_zonemgr_unrechableadd() - to add
entries to the cache. In my opinion the problem is that there is no
mechanism to remove entries from the cache except when they expire.
BIND should remove an entry for a particular master every time it
receives any communication from that master, like in
dns_zone_notifyreceive(). As you pointed out, it doesn't necesarrily
mean the master is reachable from the slave but it is worth trying to
respond to the notification, it won't hurt.

What do you think?

Take care,
Jarek