Zabbix database - mysql / postgres / noSQL

[fpaternot]

Hello there,

i am studing zabbix to deploy a very large monitoring enviroment. So far, i have 28k items and not much more than 20 hosts (core switches with a bunch of ports). I will need to deploy about 2k host in the near future, and 10k within 2 years.

After a while reading the web about the size that zabbix can grow, i have stumbled into zabbix running mysql, postgres and few mods of NoSQL. Few posts about table partitioning and clusters.

I'd like to start keeping history up to 1,5 year, and discard after that.

What would be your suggestions? Go with mod and use some NoSQL for hystory and stuff, use partitioning on tables, clusters, or what?


For NoSQL i found this pretty good and detailed paper from Miracle linux

http://www.miraclelinux.com/jp/online-service/labs/pdf/zabbix-write-performance

Thank you for your thoughts on this.

[Jason]

We've recently moved from mysql to postgres... The main reason being the support for partitioning in zabbix 2. Without partitioning then innodb tables need re-optimising on a regular basis to keep track of event deletions... With partitioning it is much neater and so far is working well.

[BDiE8VNy]

If you're willing to get the hands dirty take a look at NuoDB.
I had mentioned it a while ago by the post Any experience with NuoDB?.

This WebCast will include a live demo on Google Compute Engine and might be interesting too: The New Path to Performance. No Sharding!