I’m wondering if you are overthinking this. You said that the memory was reused when the workload increased again. Linux memory management is unintuitive. What would happen if you used a different metric, say # active connections, as your autoscaling metric? It sounds like this would behave “better”.

> On 29 Jul 2019, at 3:10 PM, Maxim Dounin <mdounin@mdounin.ru> wrote:
>
> Hello!
>
> On Mon, Jul 29, 2019 at 02:52:47PM -0400, aledbf wrote:
>
>>> on your system allocator and its settings.
>>
>> Do you have a suggestion to enable this behavior (release of memory) using a
>> particular allocator or setting?
>> Thanks!
>
> On FreeBSD and/or on any system with jemalloc(), I would expect
> memory to be returned to the OS more or less effectively.
>
> On Linux with standard glibc allocator, consider tuning
> MALLOC_MMAP_THRESHOLD_ and MALLOC_TRIM_THRESHOLD_ environment
> variables, as documented here:
>
> http://man7.org/linux/man-pages/man3/mallopt.3.html
>
> Note that you may need to use the "env" directive to
> make sure these are passed to nginx worker processes as well.
>
> --
> Maxim Dounin
> http://mdounin.ru/
> _______________________________________________
> nginx mailing list
> nginx@nginx.org
> http://mailman.nginx.org/mailman/listinfo/nginx

_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx