-
Incident report
-
Resolution: Won't fix
-
Major
-
None
-
2.4.0
-
None
-
Debian Wheezy Linux, virtual machine, 1GB ram for the zabbix server.
One can get down the monitoring server just by using the frontend and asking too much data.
To reproduce: In the frontend we login and choose "Monitoring - Latest Data" tab. Then we filter some hosts and select the checkmark on the top (which selects all items). Next you choose at the bottom "Display stacked graphs" and hit Go(50). (In my case I have 50 items. Selecting 200 items is blocked showing a message "Request-URI Too Large" which is very good)
At first glance we see the stacked graph on a 1hour basis. But clicking on "Zoom: All"(much data) causes the oom killer to be triggered on the monitoring server which will kill the mysqld service and the monitoring system is down.
Nobody can restore from this situation without an intervention of the administrator on the monitoring server to restart mysqld.
It should not be allowed that a low profiled user with almost no access to anything is able to get the complete monitoring system down.