Ok, I see，Thank you very much!
在 2020年7月8日星期三 UTC+8下午3:51:18，Stuart Clark写道：
>
> On 2020-07-08 08:44, 白洋洋 wrote:
> > Ok,thank you .But the expression
> > "container_memory_usage_bytes{container_name="prometheus"}", which is
> > a metrics of kubernetes-cadvisor, shows the memory that Prometheus
> > used is about 33G,Isn't it a little too high？Is it that most of this
> > memory is cache, and we don't need to worry about it?（The memory of
> > the computer is 48G）
>
> From your other checks it suggests most is indeed cache.
>
> As with all databases lots of memory is used to hold values stored on
> disk, to speed up queries. As mentioned this cache memory is managed by
> the kernel which will use as much as is available (totally free memory
> is a waste, so it is better to use it as cache) but throw it away if the
> memory is needed for applications.
>
> Prometheus itself does need a reasonable amount of memory to hold
> recently scraped data, handle compaction and process queries. This
> memory is managed by the application, with the amount needed dependent
> on the amount of data scraped (time series & scrape interval) and the
> number & complexity of queries.
>
> Prometheus will continue to operate if there isn't much cache memory
> available, but performance will be affected at some point - disk access
> (even SSD) is generally significantly slower than memory access. Disk IO
> usage would increase as well as CPU IO time.
>
> --
> Stuart Clark
>
--
You received this message because you are subscribed to the Google Groups
"Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to prometheus-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/prometheus-users/30000296-9997-4013-85fc-dbaef230b268o%40googlegroups.com.
