Zabbix performance - historyWriteCache

[llancellotta]

Hi, we have a large zabbix instalation with zabbix server in our on premise DC and zabbix DB in a cloud provider. Suddenly, we started to see that HistoryWriteCache is decreasing considerably fast. And performance when its closed to 0% is very poor. We think this problem is a latency problem or a database performance issue. So to leverage both, we would like to test elasticsearch implementation for the history tables and would like to ask:

1 - Elasticsearch implementation is stable in zabbix 3.4? Shoul we upgrade to zabbix 4 first?
2 - How can we migrate history to elastic? is there any tool? Or we assume that all history will be lost when we start using elasticsearch

We are using zabbix 3.4.
thanks in advance

[ingus.vilnis]

Hi and welcome to Zabbix forums!

HistoryWriteCache drop does not directly mean your DB performance is the reason and I doubt Elastic will be any help there.

First identify the root cause by analyzing all Zabbix performance graphs, pay most attention to History Syncers and Value Cache Efectiveness graph.

Also check if you don't have any heavily "spamming" items, typically logs sending hundreds of lines per second to Zabbix.

[llancellotta]

Thanks for your reply. Here is the internal processes . I have 80 History Syncer but only 30 % of use. is normal that the load is not distributed across al history syncers?

[ingus.vilnis]

80 syncers are useless in your case.

How many new values per second do you collect? Going above default 4 syncers is advised only if you have over 4000 nvps.

[llancellotta]

I have 3000 nvps actually, what do yo recommend?

[llancellotta]

Let me say that i have 60 ms between server and DB

[ingus.vilnis]

There is no way you can expect Zabbix to be working fast if you have 60ms latency between Server and DB regardless of the DB engine. You should seriously consider having both of these components in one DC.

[llancellotta]

thanks, i will move the server and let you know