|
1 | | -<!-- $PostgreSQL: pgsql/doc/src/sgml/wal.sgml,v 1.53 2008/05/02 19:52:37 tgl Exp $ --> |
| 1 | +<!-- $PostgreSQL: pgsql/doc/src/sgml/wal.sgml,v 1.54 2008/12/06 21:34:27 momjian Exp $ --> |
2 | 2 |
|
3 | 3 | <chapter id="wal"> |
4 | 4 | <title>Reliability and the Write-Ahead Log</title> |
|
135 | 135 | roll-forward recovery, also known as REDO.) |
136 | 136 | </para> |
137 | 137 |
|
| 138 | + <tip> |
| 139 | + <para> |
| 140 | + Because <acronym>WAL</acronym> restores database file |
| 141 | + contents after a crash, it is not necessary to use a |
| 142 | + journaled filesystem; in fact, journaling overhead can |
| 143 | + reduce performance. For best performance, turn off |
| 144 | + <emphasis>data</emphasis> journaling as a filesystem mount |
| 145 | + option, e.g. use <literal>data=writeback</> on Linux. |
| 146 | + Meta-data journaling (e.g. file creation and directory |
| 147 | + modification) is still desirable for faster rebooting after |
| 148 | + a crash. |
| 149 | + </para> |
| 150 | + </tip> |
| 151 | + |
| 152 | + |
138 | 153 | <para> |
139 | 154 | Using <acronym>WAL</acronym> results in a |
140 | 155 | significantly reduced number of disk writes, because only the log |
|