First there are some basic demands that must be met to make this a certified M3 solution:
There must always be, irrespective of machine type and number of disks, five folders in the root folder / naming /u01 - /u05. These folders are used for the cluster database and the M3 database.
There must be some sort of redundancy in the disk configuration. Either RAID 1, RAID 0+1 or RAID 5.
There should also be a separate file system for the online logs(pg_xlog). The amount of writing(sequential) to this file system will be extensive. Make sure it is fast. This file system can be created later on as pg_xlog will created in the cluster database from the beginning. Pg_xlog can later be linked to this file system.
Create a file system for archiving logs and backups unless you choose to use /u05(see table above). /u05 is a not very busy file system/directory.
Create directory structure for mount points. Run the following command: # mkdir –v /u01 /u02 /u03 /u04 /u05
Mount the file systems on the mount points. Use mount option ‘noatime’
Location of tablespaces, cluster database and logs for a Postgres M3 Database:
|/u02||Tablespace for M3 tables||TMVXSD|
|/u03||Tablespace for M3 indexes||TMVXSI|
|/u04||Tablespace for M3 FAT, datasets, etc..||TMVXRUNT|
|/u04||Default tablespace for M3 users/schema||TMVXTEMP|
|/u05||Tablespace for M3 archiving. Archive of transactions logs, backups||TMVXARCH|
|/pg_xlog||Postgres transaction log||N/A|