1PG_REWIND(1) PostgreSQL 11.3 Documentation PG_REWIND(1)
2
3
4
6 pg_rewind - synchronize a PostgreSQL data directory with another data
7 directory that was forked from it
8
10 pg_rewind [option...] {-D | --target-pgdata} directory
11 {--source-pgdata=directory | --source-server=connstr}
12
14 pg_rewind is a tool for synchronizing a PostgreSQL cluster with another
15 copy of the same cluster, after the clusters' timelines have diverged.
16 A typical scenario is to bring an old master server back online after
17 failover as a standby that follows the new master.
18
19 The result is equivalent to replacing the target data directory with
20 the source one. Only changed blocks from relation files are copied; all
21 other files are copied in full, including configuration files. The
22 advantage of pg_rewind over taking a new base backup, or tools like
23 rsync, is that pg_rewind does not require reading through unchanged
24 blocks in the cluster. This makes it a lot faster when the database is
25 large and only a small fraction of blocks differ between the clusters.
26
27 pg_rewind examines the timeline histories of the source and target
28 clusters to determine the point where they diverged, and expects to
29 find WAL in the target cluster's pg_wal directory reaching all the way
30 back to the point of divergence. The point of divergence can be found
31 either on the target timeline, the source timeline, or their common
32 ancestor. In the typical failover scenario where the target cluster was
33 shut down soon after the divergence, this is not a problem, but if the
34 target cluster ran for a long time after the divergence, the old WAL
35 files might no longer be present. In that case, they can be manually
36 copied from the WAL archive to the pg_wal directory, or fetched on
37 startup by configuring recovery.conf. The use of pg_rewind is not
38 limited to failover, e.g. a standby server can be promoted, run some
39 write transactions, and then rewinded to become a standby again.
40
41 When the target server is started for the first time after running
42 pg_rewind, it will go into recovery mode and replay all WAL generated
43 in the source server after the point of divergence. If some of the WAL
44 was no longer available in the source server when pg_rewind was run,
45 and therefore could not be copied by the pg_rewind session, it must be
46 made available when the target server is started. This can be done by
47 creating a recovery.conf file in the target data directory with a
48 suitable restore_command.
49
50 pg_rewind requires that the target server either has the wal_log_hints
51 option enabled in postgresql.conf or data checksums enabled when the
52 cluster was initialized with initdb. Neither of these are currently on
53 by default. full_page_writes must also be set to on, but is enabled by
54 default.
55
56 Warning
57 If pg_rewind fails while processing, then the data folder of the
58 target is likely not in a state that can be recovered. In such a
59 case, taking a new fresh backup is recommended.
60
61 pg_rewind will fail immediately if it finds files it cannot write
62 directly to. This can happen for example when the source and the
63 target server use the same file mapping for read-only SSL keys and
64 certificates. If such files are present on the target server it is
65 recommended to remove them before running pg_rewind. After doing
66 the rewind, some of those files may have been copied from the
67 source, in which case it may be necessary to remove the data copied
68 and restore back the set of links used before the rewind.
69
71 pg_rewind accepts the following command-line arguments:
72
73 -D directory
74 --target-pgdata=directory
75 This option specifies the target data directory that is
76 synchronized with the source. The target server must be shut down
77 cleanly before running pg_rewind
78
79 --source-pgdata=directory
80 Specifies the file system path to the data directory of the source
81 server to synchronize the target with. This option requires the
82 source server to be cleanly shut down.
83
84 --source-server=connstr
85 Specifies a libpq connection string to connect to the source
86 PostgreSQL server to synchronize the target with. The connection
87 must be a normal (non-replication) connection with a role having
88 sufficient permissions to execute the functions used by pg_rewind
89 on the source server (see Notes section for details) or a superuser
90 role. This option requires the source server to be running and not
91 in recovery mode.
92
93 -n
94 --dry-run
95 Do everything except actually modifying the target directory.
96
97 -P
98 --progress
99 Enables progress reporting. Turning this on will deliver an
100 approximate progress report while copying data from the source
101 cluster.
102
103 --debug
104 Print verbose debugging output that is mostly useful for developers
105 debugging pg_rewind.
106
107 -V
108 --version
109 Display version information, then exit.
110
111 -?
112 --help
113 Show help, then exit.
114
116 When --source-server option is used, pg_rewind also uses the
117 environment variables supported by libpq (see Section 34.14).
118
120 When executing pg_rewind using an online cluster as source, a role
121 having sufficient permissions to execute the functions used by
122 pg_rewind on the source cluster can be used instead of a superuser.
123 Here is how to create such a role, named rewind_user here:
124
125 CREATE USER rewind_user LOGIN;
126 GRANT EXECUTE ON function pg_catalog.pg_ls_dir(text, boolean, boolean) TO rewind_user;
127 GRANT EXECUTE ON function pg_catalog.pg_stat_file(text, boolean) TO rewind_user;
128 GRANT EXECUTE ON function pg_catalog.pg_read_binary_file(text) TO rewind_user;
129 GRANT EXECUTE ON function pg_catalog.pg_read_binary_file(text, bigint, bigint, boolean) TO rewind_user;
130
131 When executing pg_rewind using an online cluster as source which has
132 been recently promoted, it is necessary to execute a CHECKPOINT after
133 promotion so as its control file reflects up-to-date timeline
134 information, which is used by pg_rewind to check if the target cluster
135 can be rewound using the designated source cluster.
136
137 How it works
138 The basic idea is to copy all file system-level changes from the source
139 cluster to the target cluster:
140
141 1. Scan the WAL log of the target cluster, starting from the last
142 checkpoint before the point where the source cluster's timeline
143 history forked off from the target cluster. For each WAL record,
144 record each data block that was touched. This yields a list of all
145 the data blocks that were changed in the target cluster, after the
146 source cluster forked off.
147
148 2. Copy all those changed blocks from the source cluster to the target
149 cluster, either using direct file system access (--source-pgdata)
150 or SQL (--source-server).
151
152 3. Copy all other files such as pg_xact and configuration files from
153 the source cluster to the target cluster (everything except the
154 relation files). Similarly to base backups, the contents of the
155 directories pg_dynshmem/, pg_notify/, pg_replslot/, pg_serial/,
156 pg_snapshots/, pg_stat_tmp/, and pg_subtrans/ are omitted from the
157 data copied from the source cluster. Any file or directory
158 beginning with pgsql_tmp is omitted, as well as are backup_label,
159 tablespace_map, pg_internal.init, postmaster.opts and
160 postmaster.pid.
161
162 4. Apply the WAL from the source cluster, starting from the checkpoint
163 created at failover. (Strictly speaking, pg_rewind doesn't apply
164 the WAL, it just creates a backup label file that makes PostgreSQL
165 start by replaying all WAL from that checkpoint forward.)
166
167
168
169PostgreSQL 11.3 2019 PG_REWIND(1)