View Full Version : check_service_perf[ping,ip] needed
Am I missing something or is this lacking/missing? I really need the ability to have the remote client ping a particular IP address and return the ping time, would also be nice to be able to ping from the Zabbix server to an arbitrary IP address and return the time (NOT the IP address of the server, this would be used to check if the interface is reachable on the server which is on another network)
Just to say youre not the only one wanting this functionality.
I'ts also highly usefull to check the availability of VPN tunnels. For instance, with the monitored server placed in a big WAN a ping from the monitored server internally to another server placed on another wan connected via static VPN tunnels one could easily tell if the VPN is up and running.
This is quite usefull for checking the small Zyxel firewalls as they cant give this information themselves via snmp so checking the actual route is a quick and dirty way of doing it. Since those blasted firewalls can be up and running and routing just fine but slamming damn tunnels as they see fit.
Just my 2 cents.
Hello,
Zabbix is quite flexible, because you can setup your own scripts and the return value can be stored in the Zabbix database, as far as I know. With triggers, graphs, screens...
Yeah but the problem with doing your own ping check via UserParameter in the agent is that you got a max execution time that is bound to be exceted.
i use a UserParam to ping hosts from an agent
to avoid excessive execution time, i use special parameters of ping :
ping -c 30 -l 10 -f 10.0.0.1
this means i make 30 pings, 10 by 10 in parallel, as quick as possible
of course this is a bit aggressive, but you can play with parameters
the other thing is you have to be root to play with these params,
so i use sudo
the ping command above does not exceed 0.5 second over a WAN 640kbps link
cadbury> try doing that on a windows box ;)
you can just issue a ping out of say cron and store the result in a file, then have the agent read the file for your custom command. I do this a lot to avoid excessive execution time.
Same all the customs parameter are written to a file instead of being calculated directly by zabbix agent, then the command is just to take the value from the file so whatever the execution time is, zabbix always get the same time to read the value.
For example to mesure the number of httpget/sec we get the number of httpget served on apache server with a script and write the result in file, then zabbix just get the value from the file.
UserParameter's MIGHT be usable in some cases but when you're talking about having to login to each and every machine and set a different IP address/subnet etc on each one, it's a lot of work. It'd be much easier to be able to do this centrally within Zabbix.
UserParameter's MIGHT be usable in some cases but when you're talking about having to login to each and every machine and set a different IP address/subnet etc on each one, it's a lot of work. It'd be much easier to be able to do this centrally within Zabbix.
I would think that in this particuar scenario the number of servers you really need to perform this function would be less than your total number of servers. And it's pretty easy to set up, but sure.... it would be easier if it was a predefined function, lots of things would be. It'd be really nice if we had an autodiscover mode with automatic map generation too.
I just really thought you were asking for a way to get this done is all, a "how do we get there from here" scenario.