Zabbix version:
1. zabbix server : 4.0.4
2. zabbix proxy : 4.0.4
Db version:
1. mysql version: 14.14 Distrib 5.6.40
2. elasticsearch: 6.1.3
lnmp:
1. php version: 7.2
2. nginx version: AWS/1.13.6.2
3. CentOS Linux release 7.5.1804 (Core)
4. 3.10.0-862.9.1.el7.x86_64

Hello!

Please recheck if you got any frequent LLD rules:

SELECT count(*),delay from items where flags = 1 GROUP BY delay ORDER BY count(*);

This will show more precise info which discovery key is executing quite frequently.

select delay,key_,count(*) from items where flags = 1 group by delay, key_ order by delay,count(*);

Please look if you got rules installed to execute every 30s, 1m. Try to increase the time for those.

Regards,

Hi, This is the result of my inquiry. The monitoring items starting with redis are not applicable to the specific host, but the monitoring template of the co

rresponding service is imported.

This is the state of the data synchronization process during data recovery.

History index cache (it stores information about items in history cache) usage stays almost constant during the problem, but history write cache (it stores values in history cache) usage is growing. I think you have a few items with lots and lots of values. Zabbix isn't very good at handling this situation, as you see in process titles, some of history syncers are barely working, some are even idling, while most of the work is done by very few of them.

If you can access your database directly, try to figure out which items have so many values. Query for that should look something like:

select itemid,count(*) as count from history where start_time<=clock and clock<=end_time;

This may be a heavy query, so choose start_time and end_time wisely. Also do this query for history_uint, history_str, history_text and history_log.

Hi, The current historical data is only written to the index of elasticsearch and is not written to the relational database. But the problem arises from the use of elasticsearch as a historical data source from my reference to the official website documentation manual.

Sorry, I'm not that familiar with Elasticsearch. Hopefully, you got the idea where to look. At this point there is no indication of a bug in Zabbix.

Regarding screen  , the rules with 1s and 5s interval can be tricky. Please increase the time interval for these rules!

Let us know does it solve the issue.

Hi, I am sorry, I did not respond in time. I have rewritten the data source into the relational database, temporarily stopping the writing of historical data into the elasticsearch index. Now the problem has been solved.

Thank you very much for explaining the root cause. Appreciate the comeback.