Hi all,
I realise this isn't a new problem, and that various suggestions have helped fix it in the past.
I would like to just ask this question again, in the hope I can start to use graphs in a commercial way.
I have the following configuration:
1 x Zabbix server v1.8.4 (configured as the "master" within a distributed setup)
3 x Zabbix proxy servers v1.8.4 (configured to send data to the master)
The master server has 4 x 2ghz dual core processors, 20gb ram, 340gb raid 5
The number of hosts being monitored is 108, items is 3103, triggers 320, required server performance is 53.71.
The problem is that for many graphs there are gaps in the graphs (and after checking, values are missing for some intervals).
This problem ONLY affects the master server (all the proxy graphs are fine).
The values in question are:
- process availability (update interval = 60 secs)
e.g. Zabbix Agent uptime, MySQL uptime
- CPU performance (update interval = 60 secs)
e.g. processor load
- resource usage (update interval = 60 secs)
e.g. disk space usage
So from the above, I summarise the problem is either with the performance of the MySQL database or the Zabbix Agent daemon running on the master server.
Simple checks (ping check, response time etc) are all fine (also 60 sec update interval).
The gaps are regular (every 3/4 minutes), but do not follow a pattern from what I can tell. It's nothing to do with the Housekeeper process, which is at the default of every hour.
I've optimised the master server performance by fine tuning the Zabbix Server .conf file and the MySQL .conf files. Most of the values for the Server .conf file are at the default (pollers, trappers, pingers) - I've increased some of the cache values. The MySQL .conf file has been tweaked quite a bit, so that the memory usage is is around 70% of the installed memory. I've also tried different Zabbix Agent .conf values, such as the increasing the pollers (for passive checks).
The processor load consistently hovers between the 60-70% mark. As I expected, this is mostly used up by the MySQL process. The Zabbix Server and Agent processes use on average less than 1% of the CPU.
Can anyone share any tips with regard to ensuring that the Zabbix Agent daemon running on a Zabbix server works without missing values every so often?
thanks in advance,
Danny.
I realise this isn't a new problem, and that various suggestions have helped fix it in the past.
I would like to just ask this question again, in the hope I can start to use graphs in a commercial way.
I have the following configuration:
1 x Zabbix server v1.8.4 (configured as the "master" within a distributed setup)
3 x Zabbix proxy servers v1.8.4 (configured to send data to the master)
The master server has 4 x 2ghz dual core processors, 20gb ram, 340gb raid 5
The number of hosts being monitored is 108, items is 3103, triggers 320, required server performance is 53.71.
The problem is that for many graphs there are gaps in the graphs (and after checking, values are missing for some intervals).
This problem ONLY affects the master server (all the proxy graphs are fine).
The values in question are:
- process availability (update interval = 60 secs)
e.g. Zabbix Agent uptime, MySQL uptime
- CPU performance (update interval = 60 secs)
e.g. processor load
- resource usage (update interval = 60 secs)
e.g. disk space usage
So from the above, I summarise the problem is either with the performance of the MySQL database or the Zabbix Agent daemon running on the master server.
Simple checks (ping check, response time etc) are all fine (also 60 sec update interval).
The gaps are regular (every 3/4 minutes), but do not follow a pattern from what I can tell. It's nothing to do with the Housekeeper process, which is at the default of every hour.
I've optimised the master server performance by fine tuning the Zabbix Server .conf file and the MySQL .conf files. Most of the values for the Server .conf file are at the default (pollers, trappers, pingers) - I've increased some of the cache values. The MySQL .conf file has been tweaked quite a bit, so that the memory usage is is around 70% of the installed memory. I've also tried different Zabbix Agent .conf values, such as the increasing the pollers (for passive checks).
The processor load consistently hovers between the 60-70% mark. As I expected, this is mostly used up by the MySQL process. The Zabbix Server and Agent processes use on average less than 1% of the CPU.
Can anyone share any tips with regard to ensuring that the Zabbix Agent daemon running on a Zabbix server works without missing values every so often?
thanks in advance,
Danny.