strange issue with host discovery

[stuartjcarter]

We have discovery agents for our 3 IP ranges. The Zabbix client is being handled by forced install via Group Policy on the Windows 2008 domain.

Zabbix server is running on a fully patched and up to date Centos 5.x box running MySQL.

The issue is that the discovery agent is finding all the machines on our network, but none of those machines are appearing under the hosts configuration section of the control panel, so none of them can be monitored.

Help?

[untergeek]

I run the risk of sounding a fool by asking, but since you did not specify, it must be asked: Did you create any Discovery Actions to act on the discovered hosts? These can be to add hosts, link templates, etc.

Configuration -> Actions

Choose Discovery in the drop-down on the right.

[stuartjcarter]

that's a completely fair question - I am a noob at Zabbix!

The action that is specified (by my supervisor, I will add, so if it's fail it's his fault ):

Auto discovery, Windows servers. Received value like "Windows" Discovery status = "Up" Service type = "Zabbix agent". Link to template "Template_Windows" Add to group "Windows Workstations". Status: enabled.

[untergeek]

Okay, that's the Discovery Action. What's the Discovery Rule?

Configuration -> Discovery

It should contain an IP range and characteristic it's discovering.

By the way, I have noticed that using multiple discovery services can cause discovery to fail. Please show what's being used to discover so we can help review.

[stuartjcarter]

we have 3 IP ranges so there's one entry for each range.

The delay is set for 36 seconds, the checks are HTTP, ICMP Ping, Zabbix agent. Each of the 3 rules are active.

[untergeek]

I presume you mean 3600 seconds? Meaning 1 hour?

Anyway, what's the parameter you're looking for with the Zabbix Agent discovery check?

[stuartjcarter]

delay is 36 seconds, not 3600. Shall I slap my supe?

Sorry, but I don't know where to check what the agent is looking for?

[untergeek]

Yeah. Discovery that often is just hammering your db. In fact, anything less than 600 seconds (10 minutes) is way overkill. Even 600 seconds is something I'd only do for an initial discovery, then slow down to 4 to 6 hours.

Click on the discovery rule. In the resulting screen you should see the word "Checks" on the left hand side about half-way down the page, with Zabbix agent and a key inside of quotation marks. We want to know the key.

[stuartjcarter]

I have set the time to 3600 seconds - thanks for that tip.

Ah! The 3 checks are HTTP, ICMP Ping, and Zabbix agent "system.uname" - but none of them are checked... could that be the issue?

[untergeek]

No. The check boxes are only for deleting the checks.

Now that you know what you're trying to discover, it would be useful to test if the Zabbix server is actually finding the Windows servers.

From your zabbix server:

zabbix_get -s {Windows Server (fully qualified)} -k system.uname

What do you see? When I run this to a local windows box, this is what I see.

$ zabbix_get -s server.example.com -k system.uname
Windows ExampleHost 5.2.3790 Windows Server 2003 Service Pack 2 Intel IA-32

[stuartjcarter]

zabbix_get -s hostname.domain.stuff -k system.uname
Windows HOSTNAME 5.1.2600 Windows XP Service Pack 3 Intel IA-32


Thing is... the server I tried this on refused the connection. The HOSTNAME above is the name of my supe's desktop PC. So it looks like maybe the wrong template is being used - windows server instead of windows desktop?

[untergeek]

That connection refused bit is important. If the Zabbix server cannot reach the host it's supposed to monitor, it can't monitor, and not coincidentally, discover.

From the zabbix server box, try the command again. If it fails, then try:

telnet windowshost.example.com 10050

If that fails to connect, then you have to open your firewall so that the Zabbix server can reach port 10050. Also, the windows server will need to be able to reach the Zabbix server on 10051 (technically there are ways around this, but for your novice setup, you should enable this connectivity).

[stuartjcarter]

getting supe to change the firewall rules on the servers.

On the desktop side, do you have any thoughts? The agent is discovering the hosts, but they aren't appearing so that they can be managed.

[untergeek]

No, really. Your discovery action should work as soon as the firewall rules are changed. You simply had 3 discovery checks. The zabbix server is able to do HTTP and/or ICMP checks, so they appear. I'd be willing to be now if you look at Monitoring -> Discovery you have 3 columns, one for each of the 3 checks. You'll probably note that the column for the Zabbix Agent check is not green for the servers that can't be reached.

Post here again after they start to show up green in the Zabbix Agent column and see if they get added.

[stuartjcarter]

I'll post in... a little less than an hour