9 Bře
2013
9 Bře
'13
1:13
Hello Knot developers,
I'm testing the use of multiple instances of Knot on the same server.
For this simple test, I have two configurations, A.conf and B.conf, as
follows. Note that I'm bind the control interface to two different ports.
A.conf:
...
...
interfaces {
a { address A; }
}
remotes {
ctl { address 127.0.0.1; }
}
control {
listen-on { address 127.0.0.1@5553; }
allow ctl;
}
B.conf:
...
...
interfaces {
b { address B; }
}
remotes {
ctl { address 127.0.0.1; }
}
control {
listen-on { address 127.0.0.1@5554; }
allow ctl;
}
I then start both instances:
knotd -c A.conf
knotd -c B.conf
Now, if I run "knotc -c A.conf reload" then the A instance reloads, as I
expect.
However, if I run "knotc -c B.conf reload", then it is still the A
instance that reloads. The only way I can get the B instance to reload
is to run "knotc -p 5554 reload".
This is not what I expected. I expected that if I give knotc a
configuration file, then it will use the address and port numbers in
there, so that it can signal the correct instance of Knot. In
comparison, the upcoming NSD4 *does* do this. It has an "nsd-control"
command, and if I give it a configuration file as a parameter, then it
uses the control port settings from that file to signal the appropriate
instance of NSD, and not just the default one.
Would you consider getting knotc to also behave the way nsd-control
does? After all, knotc *does* read some configuration from the file, so
ideally it should also read the control section and use the appropriate
address and port numbers from there.
Regards,
Anand Buddhdev
RIPE NCC