Thread: vm.overcommit_ratio

vm.overcommit_ratio

From
PG Doc comments form
Date:
The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/16/kernel-resources.html
Description:

PostgreSQL documentation point to kernel documentation for info about
vm.overcommit_ratio. But kernel documentation slightly incorrect, so may be
write slightly more here. Kernel documentation said:

The total address space commit
for the system is not permitted to exceed swap + a
configurable amount (default is 50%) of physical RAM.

Where 50% is the default for vm.overcommit_ratio. But really not "physical
RAM" is taken in case when using hugepage memory (common practice for
postgresql). For instance I have 1TiB physical memory and 800 GiB is
reserved for hugepage (1GiB pagesize). "50% of physical RAM" will be not 500
GiB, as expected, but roughly 100GiB=0.5*(1000-800), and commit limit will
be 100GiB+swap size.

Re: vm.overcommit_ratio

From
Peter Eisentraut
Date:
On 26.12.24 08:44, PG Doc comments form wrote:
> The following documentation comment has been logged on the website:
> 
> Page: https://www.postgresql.org/docs/16/kernel-resources.html
> Description:
> 
> PostgreSQL documentation point to kernel documentation for info about
> vm.overcommit_ratio. But kernel documentation slightly incorrect, so may be
> write slightly more here.

Shouldn't we then get the kernel documentation corrected?

What is your source for this information?