Zabbix 3.04 - fping name resolution problem

[Ivo Cavalcante]

Hi all,

I've an installation of Zabbix 2.2.7 and I'm attempting an upgrade to 3.0.4 , different backend - using the upgrade scripts is not an option, I think, so I've taken the "export/import" route, which has gone well so far.

Found a problem with icmpping simple check, on a certain group of hosts, though. The item is used for checking if hosts are up or down (obviously); on previous installation (2.2.7), function would return "0" (down) in case some host's name couldn't be resolved by DNS (this kind of situation is not uncommon on my current scenario); the new installation behaves different though: I get an error (fping failed: <hostname> address not found) and the item becomes unsupported - thus, breaking trigger dependencies, since this specific one doesn't fire, keeps on OK state.

So, my questions are:
1. Has somebody else experienced the same issue?
2. If so, is this by design?
3. If so, is there an easy workaround, or I'll have to fiddle with items/triggers in order to accomplish my goal (check if a host is alive/down) ? As I said, this used to work just fine in 2.2.7 .


Thanks is advance.

[Ivo Cavalcante]

Ok guys, I searched Zabbix bug reports, found this -> ZBX-9524 and this -> ZBX-8828, indicating that this change on fping's behavior was, in fact, by design. Although understandable, this change made "icmpping" simple check kind of useless for checking hosts availability, on scenarios where you might have names not being resolved by DNS for whatever reason, since I found no way (out of the box) of checking a name "resolvability" - even though it might imply creating further itens/triggers and setting dependencies. I also can't check for specific services on all boxes - so as to replace "icmpping" for some sort of "net.[tcp | udp].service" - given differences on platform or even type of hardware, without creating another set of templates/inheritances in substitution for otherwise perfectly working ones.

Am I missing something? Is there some workaround for this situation?



I appreciate any suggestions.

[guzzijason]

Not sure if it helps your specific case, but there is an agent check for DNS:

net.dns[<ip>,name,<type>,<timeout>,<count>,<protocol>]
Example: net.dns[8.8.8.8,zabbix.com,MX,2,1]

I would love to be able to use a calculated item to solve this problem, but I'm afraid that if an "unsupported item" gets used in part of a calculated item, then the calculated item then becomes "unsupported" as well.

__Jason

[Ivo Cavalcante]

Jason,

Thanks for replying. I'm affraid I can't use net.dns key, since it's an agent type check, isn't it? Currently, I get an error before even reaching the host, exactly because of a DNS query fail (queried from Zabbix server). Furthermore, we monitor other hardware besides "plain old" computers, and we can't install an agent on these.

Am I correct?

[guzzijason]

Correct - that is an agent item only, I believe.

__Jason