Zabbix Agent Active Batch tuning client for Active mode monitoring

Zabbix Agent Active Batch tuning client for Active mode monitoring


Zabbix_server end when the number of hosts is too high, the server side to collect data, Zabbix will have serious performance problems, the main performance is as follows:


1, when the monitoring end reached a magnitude, the web operation is very card, easy to appear 502

2. Layer Break

3, the Open process (pollar) too much, even if you reduce the number of item, later to add a certain amount of machine will also have problems

Therefore, the following are mainly to two optimization direction of consideration:

1. Add proxy nodes or node mode to do distributed monitoring

2. Adjust Agentd to active mode

Since the first scenario requires an additional physical machine, the second scenario is first attempted.

One, the monitoring end zabbix_agentd.conf configuration adjustment:

Startagents=0 #客户端agent模式, only active mode, when the value is 0, the zabbix_agentd on the monitored side does not listen to the local port, so the ZABBIX_AGENTD process cannot be viewed in NETSTAT-TUNPL
Serveractive=*.*.*.* #zabbix_server的ip
Hostname=agent_host_name #重要: Host name
refreshactivechecks=1800 #被监控端到服务器获取监控项的周期
buffersize=200 #被监控端存储监控信息的空间大小
timeout=10 #超时时间
The more important parameter is that serveractive and hostname,serveractive are the specified AGENTD where the collected data is sent, and the hostname must be the same as the host name when the server side is added to the host. This allows the server side to receive the data to find the corresponding relationship.

Second, the server side of the zabbix_server.conf configuration adjustment:

startpollers=100

First, the active data collection process is reduced, originally opened to more than 700, embarrassing

starttrappers=200

And then the process of handling the data that AGENTD pushes over, it's going to be a little bigger.

Third, adjust the template

It is recommended to complete cloning a template OS Linux templates to change, the place to adjust is:

Then modify the cloned template name to templates OS Linux Active Mode

Then modify the monitor mode for each item

Select All

Then select Mass Update Batch Modification

Modify the type here for Zabbix agent (active)

Update Save

Then select the Auto-discovery policy to also modify the monitoring items inside to active mode and save.

Clicking on the item prototypes also modifies the monitor entry in the active mode and saves the template as well.

After adding the Host association template OS Linux Active Mode template can be

Add the items item for the active template so that you can customize the monitoring (such as monitoring a specific process for a cluster)