extending zabbix monitoring capability and still scale
Hello,
I use zabbix do monitor internet web servers and i do like to use it to test for things like:
- real pop account test (with not only open port test but a real USER/PASS/LIST test)
- telnet to a port and see if a string exist in the reply
- test if a web page has some predefined sentence in it
- record the time it take for a complete html page to render and not only connection time in http 1.1 way (so with domain name).
I started building some custom tests using an agent on the zabbix server that pool the host using perl scripts or shell commands (wget etc...). As all my host are in a template i fear 2 things:
- all the test are done with the same time frames and have been added with bulkloader so they all will launch the scripts at the very same moment.
- all those items need to start external programs that takes time and add load to the box and are most overkill ( wget|awk|grep etc.. to chek for "OK" on a page...) compare to the task they should do.
Then how can we scale in a situation like this ? Has anyone there such setup on a large number of host ( large to me means 2 to 3 hundreds for exemple with 34 items each most of them 5 minutes interval) ?
I am curently at 40 hosts and i wonder if i will be able to scale in this scenario. My monitoring server is a PIV 2.6Ghz, 1GB ram, scsi HD and i use zabbix 1.1b7 now.
regards,
Ghislain.
I use zabbix do monitor internet web servers and i do like to use it to test for things like:
- real pop account test (with not only open port test but a real USER/PASS/LIST test)
- telnet to a port and see if a string exist in the reply
- test if a web page has some predefined sentence in it
- record the time it take for a complete html page to render and not only connection time in http 1.1 way (so with domain name).
I started building some custom tests using an agent on the zabbix server that pool the host using perl scripts or shell commands (wget etc...). As all my host are in a template i fear 2 things:
- all the test are done with the same time frames and have been added with bulkloader so they all will launch the scripts at the very same moment.
- all those items need to start external programs that takes time and add load to the box and are most overkill ( wget|awk|grep etc.. to chek for "OK" on a page...) compare to the task they should do.
Then how can we scale in a situation like this ? Has anyone there such setup on a large number of host ( large to me means 2 to 3 hundreds for exemple with 34 items each most of them 5 minutes interval) ?
I am curently at 40 hosts and i wonder if i will be able to scale in this scenario. My monitoring server is a PIV 2.6Ghz, 1GB ram, scsi HD and i use zabbix 1.1b7 now.
regards,
Ghislain.