Postgres Pro
Downloads
Home   >   mailing lists

Thread: shared_buffers revisited

shared_buffers revisited

From
"Thalis A. Kalfigopoulos"
Date:
Another minor issue that has come to my attention is that when I define
shared_buffers=40000
I assume that it'll need a shm segment of 40000*8192=327680000 bytes
And so I set /proc/sys/kernel/shmall and shmmax accordingly only to find out that postmaster failes to start because it
requestsa shmsegment of 336404480 bytes. Where is the fallacy in my math? 

TIA,
thalis

Re: shared_buffers revisited

From
Tom Lane
Date:
"Thalis A. Kalfigopoulos" <thalis@cs.pitt.edu> writes:
> Another minor issue that has come to my attention is that when I define
> shared_buffers=40000
> I assume that it'll need a shm segment of 40000*8192=327680000 bytes
> And so I set /proc/sys/kernel/shmall and shmmax accordingly only to find out that postmaster failes to start because
itrequests a shmsegment of 336404480 bytes. Where is the fallacy in my math? 

Shared buffers are not the only things we keep in shared memory ...

            regards, tom lane