I've been playing with the web monitoring capabilities of version 1.3.4 and I must say I find them counter intuitive. I may be missing something but the only way I was able to get web monitoring to work was by associating scenarios with specific servers and using "hard coded" URLs. By specific servers I mean that associating a scenario with a template causes the test status to always remain "in progress". It also doesn't work on windows agents.
I'm used to thinking of web monitoring in a server farm context, where there are many identical servers running the same application and each of them has to be monitored separately. In this environment it is better to run the web tests from a different machine, not from an agent installed on the web servers themselves. Zabbix allows that by associating a scenario with a different agent, however the configuration then becomes a disaster because every scenario has to be created multiple times, once for each monitored server.
I see two alternatives to implementing this:
I'll be happy to know if other people need to test whole web farms and if anyone found a way to accomplish that using the web monitoring capabilities of the beta version.
Eran
I'm used to thinking of web monitoring in a server farm context, where there are many identical servers running the same application and each of them has to be monitored separately. In this environment it is better to run the web tests from a different machine, not from an agent installed on the web servers themselves. Zabbix allows that by associating a scenario with a different agent, however the configuration then becomes a disaster because every scenario has to be created multiple times, once for each monitored server.
I see two alternatives to implementing this:
- If each web monitoring scenario had a list of "targets" the scenario should be executed against. Then when the scenario is run it would be executed multiple times, once against each target in the list.
In this case the generated items, must be associated with the target hosts, and not as it is now, with the host running the tests.
In this case the scenario steps should look something like this http://{HOSTNAME}/app/index.htm - Decide that web monitoring is only executed from the Zabbix server, not from agents. In this case when a scenario is associated with a host, it wouldn't run on that host's agent, but instead would run on the server and would send the HTTP requests to the configured host. This would easily apply to templates as well. So when a scenario is associated with a template it would be executed from the server against all the hosts using that template.
This alternative limits the locations from which the tests can be run but it may be simpler to implement than the first one
I'll be happy to know if other people need to test whole web farms and if anyone found a way to accomplish that using the web monitoring capabilities of the beta version.
Eran
Comment