Pb with active agent on windows

[Cyril.ZesteInfo]

Hello,

I'm working with zabbix server 1.8.2 from the latest debian stable.
When I trying to use active agent on windows, it's not working.


My windows agent zabbix config file :
### begin conf file
Server="monitor.myserver.com"
ServerPort=10051
Hostname=zeckhome #same on zabbix server
ListenPort=10050
RefreshActiveChecks=60
DisableActive=0
DebugLevel=4
LogFile=c:\zabbix\zabbix_agentd.log

###end conf file

When I forward the 10050 port it's working as passive.
I'm using windows baseline template with all agent and triggers active from the web interface.

If I cut the forwarding on this port, Zabbix answer "host unreachable".
And no error on agent log, just telling that
"
4024:20110830:120452.781 In collect_cpustat()
3240:20110830:120453.125 In send_buffer('monitor.myserver.com','10051')
3240:20110830:120453.125 Values in the buffer 0 Max 100
3240:20110830:120453.125 Sleeping for 1 seconds
4024:20110830:120453.781 In collect_cpustat()
"
While I see, with tcpdump, packet incomming from active agent, and also in zabbix server log files.

I've tryed windows precompiled agent ver 1.8.2 and 1.8.5 with the same result.

Do I miss something, please ?

Cyril

[ghoz]

are you sure that in active mode the agent can connect to 10051 on the serve r ?

do you have other agents working in active mode ? (to eliminate a firewalling problem on the server)

[Cyril.ZesteInfo]

Hello,

There's no iptables on the zabbix server, it's dedicated stand alone one directly on the web in a data center.
Telnet , nmap , netstat told me it's open.
telnet 88.191.135.62 10051
wait for prompt.
I'm using API on to get info from my android phone, it's working good.

I've tryed active from two seperate host from two network, same result.
No error in log, like unknow hostname etc...

Reading other post aout it, telnet ok , hostname in conf file are the same, it should work, and it's working in passive mode at the moment i've forwarding port 10050 for the agent.

In debug level 4 on the agent in active mode i've only get that :
3240:20110830:120453.125 In send_buffer('monitor.myserver.com','10051')
3240:20110830:120453.125 Values in the buffer 0 Max 100
3240:20110830:120453.125 Sleeping for 1 seconds

and no data on the server.
It's a bit frustrating.

Anything else to check,please?

I will try to make active check from a linux agent tonight.

Thks.

Cyril.

[ghoz]

Reading other post aout it, telnet ok , hostname in conf file are the same, it should work, and it's working in passive mode at the moment i've forwarding port 10050 for the agent.

make sure they are exactly the same. in passive mode the parameter is not used on the agent.

In debug level 4 on the agent in active mode i've only get that :
3240:20110830:120453.125 In send_buffer('monitor.myserver.com','10051')
3240:20110830:120453.125 Values in the buffer 0 Max 100
3240:20110830:120453.125 Sleeping for 1 seconds

and no data on the server.

that line says the hagent had nothing to send this second... if you only have values in the buffer 0, it's normal the server receive nothing.

Do you see in the agent's log the retrieving of active checks ? It should happen at agent restart and every RefreshActiveChecks seconds

one other thing to note, when you make changes to the zabbix web configuration, it takes some time for the zabbix frontend to synchronize the config in the database, and then some more for the agent to get the new configuration (in active mode).

[Cyril.ZesteInfo]

hi,

Rechecking hostname in agent and web conf : exactly the same.
I didn't see in the agent's log the retrieving of active checks.
The server see packet, but simply not "use them", no error, no answer.

I've got to host in active mode test within a full day, with no update.
While there is a lot of element and triggers for both hosts and with no error in the web interface.
(I've also checked the -t parameter on agent with some element directly on agent host).

How to see the dialog with the agent and the server?

Cyril.

[ghoz]

(I've also checked the -t parameter on agent with some element directly on agent host).

the -t option work only for some internal agent items, if you really want to test, use zabbix_get on the server or on the agent (appening 127.0.0.1 to the server config parameter), but this will only test the 'passive part'.

While there is a lot of element and triggers for both hosts and with no error in the web interface.

the 'normal' triggers are fired only when data is received, you may want to add a 'nodata' part for active trigers (or you may not as it can be quite noisy in case of nework outage)


In any case , it seems that your agent in active mode doesn't seem to retrieve the list of item to check...

The server see packet, but simply not "use them", no error, no answer.

how do you know the server get's the packet ? log files ? or dump ?
in the server logfile with debuglevel 4 you should see something like

Trapper got [{
"request":"active checks",
"host":"xxx"}] len 52
In send_list_of_active_checks_json()
...
Sending [{
"response":"success",
"data":[]}]

How to see the dialog with the agent and the server?

tcpdump for the server / wireshark on the agent side :-/

[Cyril.ZesteInfo]

Solved

Hello,

After test "active only" setup agent on a linux computer, it was working by itself.

So i've looked back to the differences between windows and linux computers : I've found template.

Windows template are zabbix agent setup type by default . Looking at it, all zabbix agent elements are "active", but the type could be "zabbix agent (active)" and activited...

When i've changed all types, it worked.

Shame on me, sorry.

It was a none explicit translation from forums.

Thanks for your patience.

Have a nice day.

Cyril.