Windows 2008 DNS was in a bad state...

[frater]

Today I had a Windows 2008 server's DNS in a bad state without Zabbix detecting it.
The DNS-server was still running and even resolving records from its cache, but it wasn't able to resolve any new records.....

restarting the DNS-server resolved this error.

I would like to create a function so it can detect this state.
I thought of 2 scenarios...

The 1st is quite easy to do. I can try to resolve a record with a short TTL (like pool.ntp.org). But because I don't know in the first place why the DNS-server wasn't able to resolve it's still possible the server just returns an answer even though the record has expired.

The 2nd scenario is using a FQDN with a wildcard and asking the DNS-server a subdomain with a random string. This record can't come out of the cache (AFAIK).
My problem here is adding a random string before the FQDN


What do you guys think?

[eskytthe]

We have had similar problem with an older windows version. We have made a non-cacheable dns record witch we then test / resolve from zabbix.
BR
Erik

[frater]

Thanks for your help...

Can you elaborate on that?
What do you mean with non-cacheable?
A record with a short TTL?

[eskytthe]

Sorry frater, it was late yesterday ...

Our solution looks like your option 1. We have made a record with short TTL (less then monitor interval) but we have also placed this record at an external (other) DNS server, to force the internal server to do more the a internal lookup. Hope this make it more clearly.
Br
Erik

[frater]

Ok....

That's about the same idea I had, but I am still not 100% sure the error state that server was in will generate an error too. I can only hope so.
I will implement it and wait until a client calls with that problem.
Hopefully Zabbix detected it before that.

Do you have a low "expire" as well on the record you test?
Have you ever detected such a DNS error state?
I don't mean a complete DNS-failure, but one where it's unable to resolve new records.

[frater]

It seems net.tcp.dns() doesn't work for MS-Windows in versions < 1.9
I'm running 1.8.7 and I was planning to upgrade to the upcoming 1.9 alpha or beta. I don't want to run 1.9 alpha yet.

How are you testing DNS?
I don't like to create scripts on windows clients (Linux is something else)

[eskytthe]

Hi
The problem we see is on a windows 2000 DNS server. It can resolve cached and internal records, but stop resolving of other/external records.
We monitor the dns service from hobbit (xymon) at the moment, not zabbix as I wrongly stated fist. We transfer soon to Zabbix.
So net.tcp.dns() doesn't work for MS-Windows in versions < 1.9? Have you found this in the Jiri?