4 WAL is automatically enabled; no action is required from the
5 administrator except ensuring that the disk-space requirements for the
6 WAL files are met, and that any necessary tuning is done (see
9 WAL records are appended to the WAL files as each new record is
10 written. The insert position is described by a Log Sequence Number
11 (LSN) that is a byte offset into the WAL, increasing monotonically with
12 each new record. LSN values are returned as the datatype pg_lsn. Values
13 can be compared to calculate the volume of WAL data that separates
14 them, so they are used to measure the progress of replication and
17 WAL files are stored in the directory pg_wal under the data directory,
18 as a set of segment files, normally each 16 MB in size (but the size
19 can be changed by altering the --wal-segsize initdb option). Each
20 segment is divided into pages, normally 8 kB each (this size can be
21 changed via the --with-wal-blocksize configure option). The WAL record
22 headers are described in access/xlogrecord.h; the record content is
23 dependent on the type of event that is being logged. Segment files are
24 given ever-increasing numbers as names, starting at
25 000000010000000000000001. The numbers do not wrap, but it will take a
26 very, very long time to exhaust the available stock of numbers.
28 It is advantageous if the WAL is located on a different disk from the
29 main database files. This can be achieved by moving the pg_wal
30 directory to another location (while the server is shut down, of
31 course) and creating a symbolic link from the original location in the
32 main data directory to the new location.
34 The aim of WAL is to ensure that the log is written before database
35 records are altered, but this can be subverted by disk drives that
36 falsely report a successful write to the kernel, when in fact they have
37 only cached the data and not yet stored it on the disk. A power failure
38 in such a situation might lead to irrecoverable data corruption.
39 Administrators should try to ensure that disks holding PostgreSQL's WAL
40 files do not make such false reports. (See Section 28.1.)
42 After a checkpoint has been made and the WAL flushed, the checkpoint's
43 position is saved in the file pg_control. Therefore, at the start of
44 recovery, the server first reads pg_control and then the checkpoint
45 record; then it performs the REDO operation by scanning forward from
46 the WAL location indicated in the checkpoint record. Because the entire
47 content of data pages is saved in the WAL on the first page
48 modification after a checkpoint (assuming full_page_writes is not
49 disabled), all pages changed since the checkpoint will be restored to a
52 To deal with the case where pg_control is corrupt, we should support
53 the possibility of scanning existing WAL segments in reverse order —
54 newest to oldest — in order to find the latest checkpoint. This has not
55 been implemented yet. pg_control is small enough (less than one disk
56 page) that it is not subject to partial-write problems, and as of this
57 writing there have been no reports of database failures due solely to
58 the inability to read pg_control itself. So while it is theoretically a
59 weak spot, pg_control does not seem to be a problem in practice.