bind 9.2.1 ON rh9

[Kevin Darcy]

Mueller, Rex wrote:
> We need to look into a problem, can you point in a direction in which to
> look? 
>  
>
> We are having a problem on a box that acts as our primary DNS machine
> running Redhat 9.0 and BIND 9.2.1.. Periodically it quits replying to
> DNS requests, we clear the cache with rndc flush, sometimes, that does
> not work we have to kill -9 named processes. and restart. Sometimes it
> hangs and halts the system altogether.
>
>  
>
> I've looked in /var/log/messages there are RRSets and some lameserver
> entries but we can't seem to isolate what the problem truly is. 
>
>  
>
> The hardware is a Dell server, it'd had been working fine for a couple
> of year and as it was we'd have to rndc flush periodically (once per
> month..) now it is occuring daily. Sometimes to the point of halting the
> box. Can't say whether it's hardware or software. 
>
>  
>
> Can we put our primary DNS address (via ifconfig) to the secondary and
> take the primary offline to do hardware diagnostics? 
>
>  
>
> At this point we are baffled and need some assistance. Any insight from
> the group would be greatly appreciated.. 
>   
1. BIND 9.2.1 is pretty old at this point. I'd upgrade that, regardless 
of whether it's the fix to your problem or not.
2. If an rndc flush helps some of the time, my knee-jerk reaction is to 
say you're experiencing some sort of memory starvation issue. You didn't 
give any indication of how much memory you have in the box, how much 
named uses typically, query volume/patterns, memory usage statistics 
over a given period of time, etc. so I'm left to reckless speculation on 
that front.
3. If this problem "halts the system altogether", then, my second 
knee-jerk reaction is to say that this goes beyond a mere DNS-software 
problem, and enters the realm of OS (kernel-level) and/or hardware 
problem. Unless perhaps RedHat 9.0 is *really* bad at dealing with 
memory-starvation conditions (in which case it might just be symptomatic).
4. Can you migrate the address of your primary DNS server to another 
box, so that you can do diagnostics, without disrupting your clients? 
That's not even really a DNS question. It all depends on your 
LAN/switch/router/firewall configuration/topology. I have no idea what 
devices or paths are used between the clients and servers in your 
network, and what the configurations/rules that those devices might be 
using or not using...

                                                                         
                                 - Kevin