On Wed, Mar 23, 2022 at 02:42:06PM +0000, Shai Shapira wrote: > Hi, > > When running our application, we noticed that some processes are taking a lot of memory ( 10, 15, 20GB or so, of RSS ). > It is also reproduced when running in psql. Note that RSS can include shared_buffers read by that backend. That's a linux behavior, not specific to postgres. It's what Andres was describing here: https://www.postgresql.org/message-id/flat/20201003230149.mtd7fjsjwgii3jv7@xxxxxxxxxxxxxxxxx You have effective_cache_size = 48GB, so this seems to be working as intended. (ecc is expected to include data cached not only by postgres but by the OS page cache, too). > Memory consumption: ( of case 2, application table, using system_stats ) I'm not sure, but I guess this is just a postgres view of whatever the OS shows. > Using top: > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 15298 postgres 20 0 16.8g 1.1g 1.1g S 0.0 1.7 0:02.63 postgres > PostgreSQL 12.6 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44), 64-bit > Linux illin7504 3.10.0-1127.19.1.el7.x86_64 #1 SMP Tue Aug 11 19:12:04 EDT 2020 x86_64 x86_64 x86_64 GNU/Linux > shared_buffers | configuration file | postmaster | 2097152 | 8kB | 1024 > effective_cache_size | configuration file | user | 6291456 | 8kB | 524288 > work_mem | configuration file | user | 20480 | kB | 4096
