I need help in determining what is consuming so much memory in the cloud SQL (GCP managed) instance running with "PostgreSQL 16.6 on x86_64-pc-linux-gnu, compiled by Debian clang version 12.0.1, 64-bit".
Server spec: 16 vcpu and 128GB
Consistently, CACHE % is 59 - 60% and USAGE is around 39 - 40%. Therefore we have just around 0.5% free available memory.
By design, Linux uses all memory. What the applications don't use, Linux uses to cache data.
After all, you're paying for 128GB of RAM, so you might as well put it to good use. Since the application isn't using it, Linux uses it on your behalf..
Be concerned if USAGE is up around 90%. Be worried if it starts using swap space (if swap even exists in GCP).
I ran queries to get shared_buffers and other values.
Here are the details:
shared_buffers: 42GB
work_mem: 4MB
maintenance_work_mem: 64MB
max_connections: 5900 (however, most of the times we have under 150 concurrent sessions)
