Discovery & Template linkage

[mbrand5371]

Hi there.

I've problems automatic linking templates by discovery.

Port State on Server:
[root@slxsys07 ~]# netstat -anp | grep zabbix
tcp 0 0 0.0.0.0:10500 0.0.0.0:* LISTEN 18757/zabbix_agentd
tcp 0 0 0.0.0.0:10510 0.0.0.0:* LISTEN 18699/zabbix_server


Connection from Node:
[root@slxvbs01 ~]# telnet 10.100.194.129 10510
Trying 10.100.194.129...
Connected to zabbix1p.dmz.spielgarten.de.de (10.100.194.129).
Escape character is '^]'.
Connection closed by foreign host.
[root@slxvbs01 ~]# telnet 10.100.194.129 10500
Trying 10.100.194.129...
Connected to zabbix1p.dmz.spielgarten.de.de (10.100.194.129).
Escape character is '^]'.
Connection closed by foreign host.

Connection from Server:
[root@slxsys07 ~]# telnet 10.100.216.208 10500
Trying 10.100.216.208...
Connected to slxvzd01.referenz.spielgarten.de.de (10.100.216.208).
Escape character is '^]'.
Connection closed by foreign host.



Zabbix-Agent Log from Node:
10818:20090925:090926 zabbix_agentd started. ZABBIX 1.6.4.
10819:20090925:090926 zabbix_agentd collector started
10820:20090925:090926 zabbix_agentd listener started
10822:20090925:090926 zabbix_agentd listener started
10823:20090925:090926 zabbix_agentd listener started
10824:20090925:090926 zabbix_agentd listener started
10825:20090925:090926 zabbix_agentd listener started
10826:20090925:090926 zabbix_agentd active check started [10.100.194.129:10510]

Zabbix-Server Log:
18707:20090925:104921 Send list of active checks to [10.100.216.208] failed: host [slxvzd01.referenz.spielgarten.de.de] not found
18709:20090925:104929 Send list of active checks to [10.100.216.209] failed: host [slxvzd02.referenz.spielgarten.de.de] not found
18721:20090925:104938 Escalation canceled: Trigger [46] deleted.
18721:20090925:104938 Escalation canceled: Trigger [46] deleted.

Host lookup on Zabbix Server:
[root@slxsys07 ~]# host slxvzd01.referenz.spielgarten.de.de
slxvzd01.referenz.spielgarten.de.de has address 10.100.216.208


Uname from Node:
[root@slxvbs01 ~]# uname -a
Linux slxvbs01.referenz.spielgarten.de.de 2.6.9-89.ELsmp #1 SMP Mon Apr 20 10:34:33 EDT 2009 i686 i686 i386 GNU/Linux

Discovery Action:
Conditions (type is AND)
Service type = "ZABBIX agent"
Discovery status = "Up"
Received value like "Linux"
Actions
Link to template "LSKN_Template_Linux"



Latest Data from Node is only reporting link to "ICMP_Ping" Template.
No link to "Template_Linux".

So i need in idea, why the templates not linked.
Regards, Marcus

[mbrand5371]

Found it !

I've missed the: ZABBIX agent (10500) "system.uname"
Check

[mbrand5371]

Another issue like the first.

Discovery is up and running. Linux Agents would be linked fine.
But HP-UX Agents not

Agent response:

/opt/zabbix_agt/zabbix_agentd -t system.uname
system.uname                                  [t|HP-UX shpdev01 B.11.23 U ia64 0979926781 unlimited-user license]

Action configuration:

(A) 	Service type = "ZABBIX agent"
(B) 	Discovery status = "Up"
(C) 	Received value like "HP-UX"


Link to template "LSKN_Template_HPUX"

Communication between agent and server is open:

root@shpdev01-R:/root # telnet 10.100.194.129 10510
Trying...
Connected to 10.100.194.129.
Escape character is '^]'.

But after 12h there is no link to the HP-UX monitoring template

[dotneft]

zabbix_server.conf

StartDiscoverers=?

[mbrand5371]

@dotneft:
StartDiscoverers=6


I found it. If i start the zabbix-server it will have an listener on the configured port:

tcp        0      0 0.0.0.0:10510               0.0.0.0:*                   LISTEN      4432/zabbix_server

And at this time it will discover and link correct both Linux and HP-UX.

After some time, the listener will not belong to zabbix_server. There are often new shell processes started using new listener at port 10510, which will not be identified as zabbix_server process.

I've set the zabbix_server to debug mode.

[tob]

(running 1.6.6)

Ok, think I've found the problem here, running Debian linux anyway:

We roll with the following in zabbix_agentd.conf:

#ListenIP=127.0.0.1

So, this should mean the agent will listen on all interfaces, I really need this since the servers have a bond0 interface(for failover). I deploy the agents with puppet, having big serverfarms. setting the individual IP addresss per server in puppet is not an option, it defies the purpose of puppet

So, when the agent starts up it does:

root@xx-xxxx001:~# netstat -anp | grep zabbix
tcp6 0 0 :::10050 :::* LISTEN 19141/zabbix_agentd

Note the *tcp6*.

When I use discovery on tcp 10050 (so not zabbix-agent) the server will notice the port open though, but discovery on zabbix-agent (system.uname, really need that as well).

So, when I adjust the config file to use the bond interface I get:

ListenIP=10.X.XX.20

root@po-shellserver001:~# netstat -anp | grep zabbix
tcp 0 0 10.X.XX.20:10050 0.0.0.0:* LISTEN 9343/zabbix_agentd

Now discovery will work fine.

So, now it listen to tcp. This has a lot to do with the IPv6 support for zabbix-agent in the debian package, something we want as well.
When I add templates manually it will start working anyway, but this means discovery will not work optimal...

Devs, what shall I do, make my own debian package without v6 support?