[Nagiosplug-devel] RFC: gethostbyname instead of getaddrinfo

[Ton Voon]

Hi!

Found an issue at a customer site which I think requires a change in  
the plugins which might impact on a lot of things.

Customer setup is this:
   - host checks use check_icmp, host address is the DNS name
   - service checks use a mixture of plugins, including check_tcp

Customer had a primary DNS failure. A lot of service alerts raised.

Analysis showed that the host checks were all considered OK and the  
service checks were "(Service Check timed out)". However, check_nrpe  
service checks worked correctly.

I think this is due to the plugins using getaddrinfo (which is a DNS  
call), rather than using gethostbyname (which is an OS call). In this  
DNS setup, the OS must be caching the DNS results, hence check_icmp  
and check_nrpe working correctly. As check_tcp is using a different  
method, its results are inconsistent with the host results.

I propose changing all instances of getaddrinfo to gethostbyname in  
the plugins. This will have an added benefit of reducing the load on  
DNS servers.

I guess some plugins may specifically want to use getaddrinfo (say a  
DNS round robin web server), but I think those are the exceptions to  
the rule.

Since gethostbyname is already used by check_nrpe and check_icmp, I've  
confident of the availability of this function.

Comments?

Ton