If you've upgraded a Debian etch system running bind9 to lenny, these are some of the issues you're likely to face:
bind9 no longer allows recursive queries by default; you may need to add allow-query { any; }; to named.conf to get the previous behavior.
/etc/default/bind9 needs to contain OPTIONS="-u bind" but the etch file will not be altered during the upgrade. You must edit it by hand. This in turn will cause named to run as the user account bind, so you may also have to change the ownership or permissions of cached zone files in /etc/bind if you are a slave for any domains. (It may be best to set up a subdirectory of /etc/bind/ and give user account bind ownership and write permission, and put all your slave caches there.)
The upgrade will probably not be able to stop the named process due to changes in how rndc is configured. You may have to kill the old named process manually (assuming you don't just wait for a reboot).
If you had a logging configuration like this in named.conf:
logging { category cname { null; }; };then the new named will fail to start. This and any other obsolete logging directives must be commented out.
In order for the boot script (/etc/init.d/bind9) to be able to stop/reload/etc. the running named process, it must be able to signal it using something called "rndc". This requires the presence of two files under /etc/bind: rndc.key and rndc.conf. The key file is present after the upgrade, but the rndc.conf file may be missing. If it's missing, create it by hand with the following content:
options { default-server localhost; default-key "rndc-key"; }; include "/etc/bind/rndc.key";Ideally, you should do this before restarting named, so that it will use the correct rndc mechanism and key. (This issue may actually have arisen during a post-etch security patch, so you may already have fixed this one.)