1PG_RESETXLOG(1) PostgreSQL Server Applications PG_RESETXLOG(1)
2
6 pg_resetxlog - reset the write-ahead log and other control information
7 of a PostgreSQL database cluster
8
10 pg_resetxlog [ -f ] [ -n ] [ -ooid ] [ -x xid ] [ -e xid_epoch ]
11 [ -m mxid ] [ -O mxoff ] [ -l timelineid,fileid,seg ] datadir
12
14 pg_resetxlog clears the write-ahead log (WAL) and optionally resets
15 some other control information stored in the pg_control file. This
16 function is sometimes needed if these files have become corrupted. It
17 should be used only as a last resort, when the server will not start
18 due to such corruption.
19 After running this command, it should be possible to start the server,
21 but bear in mind that the database may contain inconsistent data due to
22 partially-committed transactions. You should immediately dump your
23 data, run initdb, and reload. After reload, check for inconsistencies
24 and repair as needed.
25 This utility can only be run by the user who installed the server,
27 because it requires read/write access to the data directory. For
28 safety reasons, you must specify the data directory on the command
29 line. pg_resetxlog does not use the environment variable PGDATA.
30 If pg_resetxlog complains that it cannot determine valid data for
32 pg_control, you can force it to proceed anyway by specifying the -f
33 (force) switch. In this case plausible values will be substituted for
34 the missing data. Most of the fields can be expected to match, but man‐
35 ual assistance may be needed for the next OID, next transaction ID and
36 epoch, next multitransaction ID and offset, WAL starting address, and
37 database locale fields. The first six of these can be set using the
38 switches discussed below. pg_resetxlog's own environment is the source
39 for its guess at the locale fields; take care that LANG and so forth
40 match the environment that initdb was run in. If you are not able to
41 determine correct values for all these fields, -f can still be used,
42 but the recovered database must be treated with even more suspicion
43 than usual: an immediate dump and reload is imperative. Do not execute
44 any data-modifying operations in the database before you dump; as any
45 such action is likely to make the corruption worse.
46 The -o, -x, -e, -m, -O, and -l switches allow the next OID, next trans‐
48 action ID, next transaction ID's epoch, next multitransaction ID, next
49 multitransaction offset, and WAL starting address values to be set man‐
50 ually. These are only needed when pg_resetxlog is unable to determine
51 appropriate values by reading pg_control. Safe values may be determined
52 as follows:
53 · A safe value for the next transaction ID (-x) may be determined by
55 looking for the numerically largest file name in the directory
56 pg_clog under the data directory, adding one, and then multiplying by
57 1048576. Note that the file names are in hexadecimal. It is usually
58 easiest to specify the switch value in hexadecimal too. For example,
59 if 0011 is the largest entry in pg_clog, -x 0x1200000 will work (five
60 trailing zeroes provide the proper multiplier).
61 · A safe value for the next multitransaction ID (-m) may be determined
63 by looking for the numerically largest file name in the directory
64 pg_multixact/offsets under the data directory, adding one, and then
65 multiplying by 65536. As above, the file names are in hexadecimal, so
66 the easiest way to do this is to specify the switch value in hexadec‐
67 imal and add four zeroes.
68 · A safe value for the next multitransaction offset (-O) may be deter‐
70 mined by looking for the numerically largest file name in the direc‐
71 tory pg_multixact/members under the data directory, adding one, and
72 then multiplying by 65536. As above, the file names are in hexadeci‐
73 mal, so the easiest way to do this is to specify the switch value in
74 hexadecimal and add four zeroes.
75 · The WAL starting address (-l) should be larger than any file name
77 currently existing in the directory pg_xlog under the data directory.
78 These names are also in hexadecimal and have three parts. The first
79 part is the ``timeline ID'' and should usually be kept the same. Do
80 not choose a value larger than 255 (0xFF) for the third part; instead
81 increment the second part and reset the third part to 0. For exam‐
82 ple, if 00000001000000320000004A is the largest entry in pg_xlog, -l
83 0x1,0x32,0x4B will work; but if the largest entry is
84 000000010000003A000000FF, choose -l 0x1,0x3B,0x0 or more.
85 · There is no comparably easy way to determine a next OID that's beyond
87 the largest one in the database, but fortunately it is not critical
88 to get the next-OID setting right.
89 · The transaction ID epoch is not actually stored anywhere in the data‐
91 base except in the field that is set by pg_resetxlog, so any value
92 will work so far as the database itself is concerned. You might need
93 to adjust this value to ensure that replication systems such as
94 Slony-I work correctly — if so, an appropriate value should be
95 obtainable from the state of the downstream replicated database.
96 The -n (no operation) switch instructs pg_resetxlog to print the values
98 reconstructed from pg_control and then exit without modifying anything.
99 This is mainly a debugging tool, but may be useful as a sanity check
100 before allowing pg_resetxlog to proceed for real.
101
103 This command must not be used when the server is running. pg_resetxlog
104 will refuse to start up if it finds a server lock file in the data
105 directory. If the server crashed then a lock file may have been left
106 behind; in that case you can remove the lock file to allow pg_resetxlog
107 to run. But before you do so, make doubly certain that there is no
108 server process still alive.
109Application 2008-06-08 PG_RESETXLOG(1)