relationshipe between values and request.values of value cache

[Alberto Tagliasacchi]

Hi,
working with Zabbix 6.0.
I'm trying to figure how to deal with value cache and I'm wondering why I have big mismatch between "values" and "request.values" in the output of the following command
zabbix_server -R diaginfo=valuecache

Top.values:
itemid:2506634 values:69208 request.values:21
itemid:2506641 values:69208 request.values:21
itemid:49833 values:304 request.values:289
itemid:50009 values:304 request.values:289
itemid:64619 values:304 request.values:289
itemid:27518 values:304 request.values:289
itemid:64618 values:304 request.values:289
..
As far as I can understand:
values: number of measures of the item in cache
request.values: number of measures asked and found in the cache

I'm wondering why there are 2 items (2506634 and 2506641) with around 69000 values whereas request values are only 21.

The 2 ((2506634 and 2506641) items are used in a trigger that computes avg on such items on a 10 min base so it needs 21 values (from o to 600 every 30 sec) exactly those retrieved form cache
The 2 items are measured every 30 seconds, so I have 2880 new values per day, and the "values counter" increases each day exactrly of the same amount.

I'm wondering why the value cache are mantained so long and not deleted: the trigger looks like to use only 21 values, why do Zabbix need to save in cache, increasing then every day, and not removing from cache?
Where can I search for details?

Thanks in advance

Alberto

[Markku]

In my home installation I see similar effect, and the value cache utilization is like 5%. Let's put it the other way around: why should the code try to purge values from the cache when there is lots of space available?

How is your value cache utilization?

Markku

[Alberto Tagliasacchi]

Hi Markku!
in my case, the value cache utilization is 48%. I read it should be less than 25% and before increasing it, I was trying to understand where/what was used for and being fill of
Thanks to You.
Alberto

[Markku]

I've always thought a healthy appproximate max utilization is 75%.

But you are correct that the algorithm behind the cache population is interesting.

Markku

[Alberto Tagliasacchi]

Thanks for your attention Markku.