Problem with Auto Registration in Zabbix 3.4 (all 3.4, clients and server!)

[The Thor]

Hi all,

I have a problem with Zabbix Auto Registration actions.

I have a set of compute nodes I wish to register in my zabbix server.

First some definitions:
a) The nodes are nodeXX.sub.dom.tld where 00<=XX<=99
b) The nodes have IP addresses 10.10.0.YY where YY=XX+10
c) The 10.10/24 network goes through the 10.7/24 network (over 10.10.0.1/10.7.0.1 )
d) The 10.7/24 network conencts to the 10.1/24 network where the zabbix server (10.1.0.1) lives.
e) Communication from the zabbix server to node49 (our test subj in this post) is possible as proven by "echo system.uptime | nc node49.sub.dom.tld 10050" from the zabbix server, and get "ZBXD1407237" back.
f) Communication from node49 is also possible as proven by -> echo "ZBXDi{"request":"sender data","data":[{"host":"node49.sub.dom.tld","key":"system.uptime" ,"value":"1"}]}" | nc 10.1.0.1 10051 <- and receive "OK" back.
g) The ACTION is defined as such:
cond 1) Host metadata like: Linux
cond 2) Host metadata like: 00000000-0000-0000-0000-000000000000
cond 3) Host name like: node*.sub.dom.tld
and
oper 1) Send message to user groups: Zabbix administrators via all media
oper 2) Link to templates: Template OS Linux Cluster
h) Template OS Linux Cluster belongs to the group Cluster Nodes and incorporates the Template OS Linux template.

So, the all-important tl;dr part behind us, the message I get in the Zabbix server log is:
3814:20180521:222219.923 cannot send list of active checks to "10.7.0.1": host [node49.sub.dom.tld] not found

But here we already know that the request came from node49, correctly identifying itself as node49.sub.dom.tld and resolvable as 10.10.0.59.

So, WHY does 10.7.0.1 come there as the machine targeted for the active tests - and why is the auto registration failing?

I am, at the lack of a better word - flabbergasted, and would be immensely grateful if someone could help out

[aigars.kadikis]

Hi, Thor,

Am I correctly understand that every node gets assigned only one template 'Template OS Linux Cluster' which incorporates with 'Template OS Linux template'?
So you have active items in one of those templates? Do you have passive items too, do they work?

Please share zabbix_agentd.conf from one of your nodes.

grep -v "^$\|^#" /etc/zabbix/zabbix_agentd.conf

A screenshot from host configuration also would be nice related to "cannot send list of active checks to".

Regards,

[The Thor]

Ah! I thought it hadn't been posted here since I tried twice to post but saw no result. Thus I didn't think about updating my alledgedly nonexistant post

I did find the solution.

Apparently, a service person that services the cluster had the bright idea of routing/forwarding the network - without turning off the pre-existing NAT rule. Thus, traffic was correctly routed from server to node, but NATed from node to server. Which explains the ghost IP.

Never mix NAT and full routers children, they rot your teeth -_-

Thanks anyways, of course Zabbix auto-registration works perfectly with that pesky thing out of the way