Insufficient Housekeeping ?

[Mattéo45]

Hello everyone,

I'm coming to you because I've noticed that housekeeping doesn't delete enough data or doesn't have enough time to delete everything. Indeed, I have already increased the disk size of my VM at the end of 2024 and is already excessively increased. Attached is a snapshot of disk space since January 1, 2025 :


I've also seen that the housekeeper works well: “cat /var/log/zabbix/zabbix_server.log | grep housekeeper” result of the attached command :

Can you interpret the result? Do you know the best practices for regulating this parameter to normal? Here's some additional information:
zabbix-server.conf :
HousekeepingFrequency=1
MaxHousekeeperDelete=50000
I've already reduced the retention of some history, but that didn't change anything.

If you need more information about my configuration, don't hesitate to contact me.

Thanks for your help
Mattéo

[PavelZ]

Don't understand why you don't show data that allows you to estimate the scale of the installation?
For example, the size of the database in bytes and number of values, NVPS

It is obvious that with a duration of 25000s and a launch frequency of once per hour (HousekeepingFrequency=1), this process does not have time.

Deleting historical data is not a strong point in Zabbix. Therefore, solutions like mysql partitioning and postgresql timescale are popular. These solutions delete data in large blocks and quickly

[Mattéo45]

Hello

The DB size is 509Gb today. The number of new values per second is 783.69 with 389 hosts enabled (and is likely to increase again in the coming months). That's 67780 enabled elements and 40874 triggers, and we use the default Zabbix templates most of the time. I've created some myself to perform bandwidth calculations on switch links and availability calculations (ping over 30 days) to create indicators for the top management.

Timescale is not installed on my VM. I'll see if I can find a tutorial to implement it on an existing installation (I don't know much about it) and see if it improves the situation.

[cyber]

Seems like it is a all-in-one instance? Time to split it up, at least move DB to separate host(s). Give it a lot more power than your server/frontend has. Implement partitions/timescale. As it was already said, it improves a lot. Running a housekeeper for 6h is as good as not running it at all..:P Move all polling duties to proxies....
How come you have so many items/triggers to delete all the time?

[Mattéo45]

Yes, it's an all-in-one instance. Before trying to split the Zabbix Database and App on two servers, I'll try the TimescaleDB installation.

More info in my answer above about items and triggers.

Also for the record, server performances is fine, except the ever-increasing disk space...

[cyber]

Mattéo45 Some info about implementing TS.. https://www.zabbix.com/events/zabbix...nda#day2_17:15 You can do it with shorter time (with downtime)...