PITR (Point-in-Time Recovery) backup is one of the vital parts of a PostgreSQL environment. It is involves two kinds of processes - base backup of the database cluster and WAL archiving.
The most common way of configuring it can be described with these steps:
Creates an archived base backup in a current date named directory
in PITR_ARCHIVE_DIR, removes outdated base backup directories, in
accordance with PITR_KEEP_DAYS, from PITR_ARCHIVE_DIR, and cleans
WAL files, that are older than the oldest kept base backup, from
PITR_WAL_ARCHIVE_DIR. If PITR_LOCAL_DIR is specified, it creates
the base backup in this directory first and then RSYNC it to
PITR_ARCHIVE_DIR. If PITR_WAL is set to true it starts a process
of streaming WAL to PITR_WAL_ARCHIVE_DIR using PGRECEIVEXLOG. The
streaming process is made for running as a cron job and exits
normally if another instance is running with the same
PITR_WAL_RECEIVER_LOCK_FILE.
test -z "$PITR_WAL" && PITR_WAL=false
PITR_LOCAL_DIR=
PITR_ARCHIVE_DIR='/mnt/archive/basebackups'
PITR_WAL_ARCHIVE_DIR='/mnt/archive/wal'
test -z "$PITR_WAL_RECEIVER_LOCK_FILE" && \
PITR_WAL_RECEIVER_LOCK_FILE='/tmp/wal_receiver.'$(
echo $PITR_WAL_ARCHIVE_DIR | sed 's/\//_/g'
)"-$HOST-$PORT-$USER"
PITR_KEEP_BACKUPS=2