Shulou(Shulou.com)06/02 Report--

From: http://www.xiaomastack.com/2014/10/10/zabbix02/

Zabbix is a high-performance distributed monitoring and alarm system. For example, the common home desktop configuration processor I5-3470, memory 4GB1600MHz, hard disk 7200rpm can monitor about 1000 HOST, yes, that's right, Zabbix can achieve such high performance. There will be a small bottleneck when Zabbix is running for a long time, and the biggest bottleneck is the database. How to solve the bottleneck, provide some possible ways.

First, zabbix-agent is also installed on zabbix-server, and the Template App Zabbix Server template is enabled to monitor the parameters within the zabbix-server service. After enabling the template, the corresponding monitoring item item, graphic graph and trigger trigger will be added.

There will be a lot more monitoring items like this.

Newly added graphics

Triggers like this will also increase a lot.

Then when there is a performance bottleneck in Zabbix Server, it will issue an alarm and change the configuration parameters of zabbix-server to adapt to the current working environment according to the alarm entry of the trigger.

The following is an analysis of possible scenarios

1. On Zabbix data gathering process busy issues (performance monitoring of zabbix-server processes when server collects data)

The monitoring items are as follows:

For example, the discoverer monitoring item alarm: "Zabbix discoverer processes 75 busy" actually check the monitoring chart or monitoring value of this item may reach 100% at this time.

The solution. Edit the configuration file / usr/local/zabbix/etc/zabbix-server.conf, increase the StartDiscoverers parameter, default to 1, and increase it to the appropriate value, such as 5. 0. Then restart zabbix-server. If the utilization of the enabled discoverer service process continues to soar, it will continue to increase the number of discoverer startup processes.

1 2 3 [root@Zabbix-Server] # vim / usr/local/zabbix/etc/zabbix-server.conf StartDiscoverers=5 [root@Zabbix-Server] # / etc/init.d/zabbix_server restart

For example, the ipmi pinger monitor alarm: "Zabbix icmp pinger processes more than 75 busy", also change the server configuration file, increase the parameter StartPingers, from the default 1 to the appropriate value, such as 5. And then restart zabbix-server.

1 2 3 [root@Zabbix-Server] # vim / usr/local/zabbix/etc/zabbix-server.conf StartPingers=5 [root@Zabbix-Server] # / etc/init.d/zabbix_server restart

The parameters and default values of each monitoring item in this process load monitoring chart are as follows, and the corresponding values are adjusted when a process is overloaded:

1 2 3 4 5 6 7 8 9 10 11Zabbix busy trapper processes, in% StartTrappers=5 Zabbix busy poller processes, in% StartPollers=5 Zabbix busy ipmi poller processes, in% StartIPMIPollers=0 Zabbix busy discoverer processes, in% StartDiscoverers=1 Zabbix busy icmp pinger processes, in% StartPingers=1 Zabbix busy http poller processes, in% StartHTTPPollers=1 Zabbix busy proxy poller processes, in% StartProxyPollers=1 Zabbix busy unreachable poller processes, in% StartPollersUnreachable=1 Zabbix busy java poller processes, in% StartJavaPollers=0 Zabbix busy snmp trapper processes, in% StartSNMPTrapper=0 Zabbix busy vmware collector processes, in% StartVMwareCollectors=0

2. About Zabbix cache usage (monitoring of free values of all kinds of server cached data)

The monitoring items are as follows:

The parameters and default values of the corresponding configuration files for each monitoring item in this cache monitoring chart are as follows. Increase the capacity to the appropriate value when a certain buffer is insufficient:

1 2 3 45 6Zabbix-server: Zabbix trend write cache,% free TrendCacheSize=4M Zabbix-server: Zabbix configuration cache,% free CacheSize=8M Zabbix-server: Zabbix text write cache,% free HistoryTextCacheSize=16M Zabbix-server: Zabbix history write cache,% free HistoryCacheSize=8M Zabbix-server: Zabbix value cache,% free ValueCacheSize=8M Zabbix-server: Zabbix vmware cache,% free VMwareCacheSize=8M

3. The monitoring item Item should set an appropriate data sampling interval interval, which should not be less than 1 minute. For quantities that remain the same for a long time, such as total memory size, total disk size, etc., the interval should be as large as possible, such as one day (1D) or one week (1w).

4. There are too many hosts, how else can we only split the architecture, adopt a distributed architecture to reduce the pressure on zabbix server, and share these pressures on proxy.

5. Historical data is the main reason why the database is full. Reduce the preservation time of historical data, the default is 90 days, can be adjusted to 7 days or less, rest assured that zabbix also keeps trend data, macro historical data will not be lost.

6. Split large tables of history type (history, history_uint, etc.). Turn off housekeeper to prohibit the automatic regular cleaning of historical data, because it seems very painful for the database, especially for the InnoDB engine big data.

Reprint please indicate the source: http://www.xiaomastack.com/2014/10/10/zabbix02/ thank you!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.