auto-registering hosts on disparate networks

[tolland]

Hi,

I would like to use the zabbix agent for monitoring our EC2 instances, and our build currently includes the zabbix_agentd which starts up fine.

However I was looking to configure autodiscovery, and it appears to have been removed at some point. I looked at the discovery options (in 1.4.4) and there doesn't seem to be anything suitable for what I am after. Obviously I don't want to be scanning the entire Amazon EC2 public IP range for zabbix agents, and scripting something up to use the Import XML function looks like I would need to store and pass an admin password from my EC2 hosts, which I would prefer not to do.

Any suggestions?

Many Thanks,

T

[lamont]

Hi,

I would like to use the zabbix agent for monitoring our EC2 instances, and our build currently includes the zabbix_agentd which starts up fine.

However I was looking to configure autodiscovery, and it appears to have been removed at some point. I looked at the discovery options (in 1.4.4) and there doesn't seem to be anything suitable for what I am after. Obviously I don't want to be scanning the entire Amazon EC2 public IP range for zabbix agents, and scripting something up to use the Import XML function looks like I would need to store and pass an admin password from my EC2 hosts, which I would prefer not to do.

Any suggestions?

Many Thanks,

T

In your build, have a script which on boot, and periodically once a day reports to some centralized known service that the host is alive and what IP interfaces are configured on it. Then you will have a directory of all your EC2 hosts and which IP addresses they are on (although can't you get this from EC2 somewhere? Amazon has that information stored in a database associated with your account, can you just grab it from there?).

Once you've got that information there's several ways to skin the cat. One way is to write some SQL which will insert /32 entries into drules and dchecks (just be sure to increment druleid and dcheckid values in the ids tables). Zabbix will then scan those IP addresses and do auto-discovery on them. Once this is running as an automated processes, then you can just use the discovery interface in Zabbix to control how you discover what kind of host it is and what kind of monitoring gets applied to it.

[Alexei]

FYI Support of auto-registration will be restored in one of 1.6.x releases, quite likely that in 1.6.

[tolland]

Once you've got that information there's several ways to skin the cat. One way is to write some SQL which will insert /32 entries into drules and dchecks (just be sure to increment druleid and dcheckid values in the ids tables). Zabbix will then scan those IP addresses and do auto-discovery on them. Once this is running as an automated processes, then you can just use the discovery interface in Zabbix to control how you discover what kind of host it is and what kind of monitoring gets applied to it.

I looked at writing the SQL directly, but I couldn't find any documentation on the DB schema and the mysql binlog viewer showed a whole bunch of records being written, and I don't really understand how the templates worked so I sacked that off.

Then I moved onto trying to pull the relevant classes out of the php frontend files, and call the "function add_host($host,$port,$status,$useip,$dns,$ip,$temp lates,$newgroup,$groups)"
function from a script. But I ran into all sorts of problems with understanding the include files, and web context/environments settings that I couldn't replicate in a scripted version.

In the end, the simpliest being the best, I chucked a short bash script together which replicates logging in, and sending the required POST variables to hosts.php - It goes a little something like this;

#!/bin/bash

#called as add_zabbix_host.sh $public_hostname $instance_id $public_ipv4

DNS=$1
HOST=$2
IP=$3

USERNAME=myuseretc
PASSWORD=mypasss

wget --keep-session-cookies --save-headers --save-cookies /tmp/cookies -O- --post-data="form=1&form_refresh=1&name=$USERNAME&password=$PASSWORD&enter=Enter" http://dcdev01.osti.local/zabbix/index.php

wget --keep-session-cookies --save-headers --load-cookies /tmp/cookies -O- --post-data="form=Create+Host&form_refresh=5&config=0&groupid=0&devicetype=&name=&os=&serialno=&tag=&macaddress=&hardware=&software=&contact=&location=&notes=&host=$HOST&groups%5B%5D=15&newgroup=&dns=$DNS&ip=$IP&useip=0&port=10050&status=0&save=Save&templates%5B10001%5D=Template_Linux"  http://dcdev01.osti.local/zabbix/hosts.php 

[nelsonab]

A rough outline of the schema can be found in the 1.1 documentation. It has changed a little since then but it does give you a good start.

[lamont]

Yeah, you're trying to be a bit aggressive about it. I would use Zabbix's auto-discovery feature but feed the auto-discovery a set of IP addresses that you know are all of your hosts. Then all you have to do is insert a row into dchecks and a row into drules for the IP address.

Then zabbix will be able to auto-discover the agent on that IP address. Then you setup your discovery actions to associate the discovered hosts with groups, templates, etc.

And you might need to clean out dhosts and dservices to force a re-discovery ("delete from dhosts; delete from dservices;") I've found that's a convenient way to keep templates and groups associated with hosts in a repeatable way by keeping my discovery action rules updated and then forcing re-discovery.

Anyway this keeps the SQL fiddling down to just the dchecks and drules tables. All the addition of hosts, association of templates and groups and items and all that other junk happens through normal Zabbix code as a function of the discovery event.

I looked at writing the SQL directly, but I couldn't find any documentation on the DB schema and the mysql binlog viewer showed a whole bunch of records being written, and I don't really understand how the templates worked so I sacked that off.

Then I moved onto trying to pull the relevant classes out of the php frontend files, and call the "function add_host($host,$port,$status,$useip,$dns,$ip,$temp lates,$newgroup,$groups)"
function from a script. But I ran into all sorts of problems with understanding the include files, and web context/environments settings that I couldn't replicate in a scripted version.

In the end, the simpliest being the best, I chucked a short bash script together which replicates logging in, and sending the required POST variables to hosts.php - It goes a little something like this;

#!/bin/bash

#called as add_zabbix_host.sh $public_hostname $instance_id $public_ipv4

DNS=$1
HOST=$2
IP=$3

USERNAME=myuseretc
PASSWORD=mypasss

wget --keep-session-cookies --save-headers --save-cookies /tmp/cookies -O- --post-data="form=1&form_refresh=1&name=$USERNAME&password=$PASSWORD&enter=Enter" http://dcdev01.osti.local/zabbix/index.php

wget --keep-session-cookies --save-headers --load-cookies /tmp/cookies -O- --post-data="form=Create+Host&form_refresh=5&config=0&groupid=0&devicetype=&name=&os=&serialno=&tag=&macaddress=&hardware=&software=&contact=&location=&notes=&host=$HOST&groups%5B%5D=15&newgroup=&dns=$DNS&ip=$IP&useip=0&port=10050&status=0&save=Save&templates%5B10001%5D=Template_Linux"  http://dcdev01.osti.local/zabbix/hosts.php 

[lamont]

FYI Support of auto-registration will be restored in one of 1.6.x releases, quite likely that in 1.6.

Oh, I see discovery got broken? I'm using 1.4.4 and so far its been working fine with addition of my dont-trust-DNS patch.

[thunderbolt]

Alexei,

Any update on whether auto-registration will make it into the upcoming 1.6 release that I saw is scheduled for the 18th?

I'd really like to use it instead of writing a bunch of custom scripts to make discovery work for remote systems on different networks.

Thanks,
Ron

[pierre-hoffmann]

Hi,

i'm also use wget to auto-create host in Zabbix from ksh shell scripts;
it exists a cleanner method to do this or not.

Nothing in 1.6 or greater ??

Regards,