루트엔 설치 가급적 하지마시고 별도의 공간에 파일시스템을 만들고 해당 명칭으로 링크를 걸어주세요..
2024년 8월 30일 (금) 오전 11:34, Sam Stearns <sam.stearns@xxxxxxx>님이 작성:
Thanks, Rui!On Fri, Aug 30, 2024 at 11:52 AM Rui DeSousa <rui.desousa@xxxxxxxxxx> wrote:On Aug 29, 2024, at 2:12 PM, Henry Ashu <henry.ashu@xxxxxxx> wrote:We have installed our postgresql software in this location:[postgres@testdbapql01 data]$ pwd
/var/lib/pgsql/16/data
[postgres@testbapql01 data]$ df -h .
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/rhel-root 28G 4.5G 23G 17% /
[postgres@testdbapql01 data]$Does this look correct? We are thinking that it should not be installed on the root filesystem.Yes, it is better to have more mount points than not. Root should be its own mount point; OSs do not like when the root volume is full thus it is normally a very small filesystem.Disclaimer: I’ve seen servers with a single mount point /. A lot of devops servers are setup with a single mount point for the entire OS. Seems like that is the case with your system. The reason to have multiple filesystem is space management, mount point options, and filesystem corruption. I haven’t had the need to run fsck in decades until last month. An xfs volume was corrupted and even fsck couldn’t fix it; had to resort to a filesystem snapshot.I normally have three;1. /var/log — for log files via syslog. You don’t want a run away application to full up the log directory and crash the server.2. $PGDATA — data volume3. $PGDATA/pg_wal — Normally, I’m using ZFS with different filesystem properties. It’s also a good idea to manage the space separately.As far as absolute mount points that’s a preference; my current gig is using the following with the instance names as a directory under those mount points./pg_data/pg_wal--