本节内容来源于PGer的一个问题:
Q:
为什么checkpointer后台进程使用这么多内存?


[pg12@localhost ~]$ ps -aux|grep postgres
pg12      1506  0.0  1.8 753156 34704 ?        Ss   10:30   0:00 /appdb/xdb/pg12beta1/bin/postgres
pg12      1507  0.0  0.0 187304  1472 ?        Ss   10:30   0:00 postgres: logger   
pg12      1509  0.0 24.1 754120 452672 ?       Ss   10:30   0:04 postgres: checkpointer   
pg12      1510  0.0 28.4 753292 533184 ?       Ss   10:30   0:08 postgres: background writer   
pg12      1511  0.0  0.9 753260 18012 ?        Ss   10:30   0:11 postgres: walwriter   
pg12      1513  0.0  0.0 189420  1448 ?        Ss   10:30   0:00 postgres: archiver   last was 00000001000000020000005C
pg12      1514  0.0  0.0 189556  1324 ?        Ss   10:30   0:00 postgres: stats collector   
pg12      1515  0.0  0.1 753688  1984 ?        Ss   10:30   0:00 postgres: logical replication launcher  
pg12      2281  0.0  0.0 112660   972 pts/3    S+   16:38   0:00 grep --color=auto postgres
[pg12@localhost ~]$ grep 'shared' $PGDATA/postgresql.conf 
shared_buffers = 512MB            # min 128kB
#shared_memory_type = mmap        # the default is the first option
dynamic_shared_memory_type = posix    # the default is the first option
#wal_buffers = -1            # min 32kB, -1 sets based on shared_buffers
#shared_preload_libraries = ''    # (change requires restart)

比如上例,共享内存设置为512MB,在2G内存的机器上,checkpointer和background writer进程占用了差不多25%的内存大小.

A:

resident” RAM of a postgres subprocess is often just be the fraction of shared_buffers it’s read/written. checkpointer must necessarily read all dirty pages from s-b and write out to disk (by way of page cache), so that’s why its RSS is nearly 32GB. And the recovery process is continuously writing into s-b.

原因是checkpointer和bg writer需要从shared memory中读取数据(从s-b到page cache)并刷到磁盘上,因此给人占用大量内存的假象.

参考资料
Re: checkpoint and recovering process use too much memory

PostgreSQL DBA(56) - Why does checkpointer use so much memory