4 Logical replication behaves similarly to normal DML operations in that
5 the data will be updated even if it was changed locally on the
6 subscriber node. If incoming data violates any constraints the
7 replication will stop. This is referred to as a conflict. When
8 replicating UPDATE or DELETE operations, missing data is also
9 considered as a conflict, but does not result in an error and such
10 operations will simply be skipped.
12 Additional logging is triggered, and the conflict statistics are
13 collected (displayed in the pg_stat_subscription_stats view) in the
14 following conflict cases:
17 Inserting a row that violates a NOT DEFERRABLE unique
18 constraint. Note that to log the origin and commit timestamp
19 details of the conflicting key, track_commit_timestamp should be
20 enabled on the subscriber. In this case, an error will be raised
21 until the conflict is resolved manually.
23 update_origin_differs #
24 Updating a row that was previously modified by another origin.
25 Note that this conflict can only be detected when
26 track_commit_timestamp is enabled on the subscriber. Currently,
27 the update is always applied regardless of the origin of the
31 The updated value of a row violates a NOT DEFERRABLE unique
32 constraint. Note that to log the origin and commit timestamp
33 details of the conflicting key, track_commit_timestamp should be
34 enabled on the subscriber. In this case, an error will be raised
35 until the conflict is resolved manually. Note that when updating
36 a partitioned table, if the updated row value satisfies another
37 partition constraint resulting in the row being inserted into a
38 new partition, the insert_exists conflict may arise if the new
39 row violates a NOT DEFERRABLE unique constraint.
42 The row to be updated was not found. The update will simply be
43 skipped in this scenario.
45 delete_origin_differs #
46 Deleting a row that was previously modified by another origin.
47 Note that this conflict can only be detected when
48 track_commit_timestamp is enabled on the subscriber. Currently,
49 the delete is always applied regardless of the origin of the
53 The row to be deleted was not found. The delete will simply be
54 skipped in this scenario.
56 multiple_unique_conflicts #
57 Inserting or updating a row violates multiple NOT DEFERRABLE
58 unique constraints. Note that to log the origin and commit
59 timestamp details of conflicting keys, ensure that
60 track_commit_timestamp is enabled on the subscriber. In this
61 case, an error will be raised until the conflict is resolved
64 Note that there are other conflict scenarios, such as exclusion
65 constraint violations. Currently, we do not provide additional details
68 The log format for logical replication conflicts is as follows:
69 LOG: conflict detected on relation "schemaname.tablename": conflict=conflict_ty
71 DETAIL: detailed_explanation.
72 {detail_values [; ... ]}.
74 where detail_values is one of:
76 Key (column_name [, ...])=(column_value [, ...])
77 existing local row [(column_name [, ...])=](column_value [, ...])
78 remote row [(column_name [, ...])=](column_value [, ...])
79 replica identity {(column_name [, ...])=(column_value [, ...]) | full [(colu
80 mn_name [, ...])=](column_value [, ...])}
82 The log provides the following information:
86 + schemaname.tablename identifies the local relation involved in
88 + conflict_type is the type of conflict that occurred (e.g.,
89 insert_exists, update_exists).
93 + detailed_explanation includes the origin, transaction ID, and
94 commit timestamp of the transaction that modified the existing
95 local row, if available.
96 + The Key section includes the key values of the local row that
97 violated a unique constraint for insert_exists, update_exists
98 or multiple_unique_conflicts conflicts.
99 + The existing local row section includes the local row if its
100 origin differs from the remote row for update_origin_differs
101 or delete_origin_differs conflicts, or if the key value
102 conflicts with the remote row for insert_exists, update_exists
103 or multiple_unique_conflicts conflicts.
104 + The remote row section includes the new row from the remote
105 insert or update operation that caused the conflict. Note that
106 for an update operation, the column value of the new row will
107 be null if the value is unchanged and toasted.
108 + The replica identity section includes the replica identity key
109 values that were used to search for the existing local row to
110 be updated or deleted. This may include the full row value if
111 the local relation is marked with REPLICA IDENTITY FULL.
112 + column_name is the column name. For existing local row, remote
113 row, and replica identity full cases, column names are logged
114 only if the user lacks the privilege to access all columns of
115 the table. If column names are present, they appear in the
116 same order as the corresponding column values.
117 + column_value is the column value. The large column values are
118 truncated to 64 bytes.
119 + Note that in case of multiple_unique_conflicts conflict,
120 multiple detailed_explanation and detail_values lines will be
121 generated, each detailing the conflict information associated
122 with distinct unique constraints.
124 Logical replication operations are performed with the privileges of the
125 role which owns the subscription. Permissions failures on target tables
126 will cause replication conflicts, as will enabled row-level security on
127 target tables that the subscription owner is subject to, without regard
128 to whether any policy would ordinarily reject the INSERT, UPDATE,
129 DELETE or TRUNCATE which is being replicated. This restriction on
130 row-level security may be lifted in a future version of PostgreSQL.
132 A conflict that produces an error will stop the replication; it must be
133 resolved manually by the user. Details about the conflict can be found
134 in the subscriber's server log.
136 The resolution can be done either by changing data or permissions on
137 the subscriber so that it does not conflict with the incoming change or
138 by skipping the transaction that conflicts with the existing data. When
139 a conflict produces an error, the replication won't proceed, and the
140 logical replication worker will emit the following kind of message to
141 the subscriber's server log:
142 ERROR: conflict detected on relation "public.test": conflict=insert_exists
143 DETAIL: Key already exists in unique index "t_pkey", which was modified locally
144 in transaction 740 at 2024-06-26 10:47:04.727375+08.
145 Key (c)=(1); existing local row (1, 'local'); remote row (1, 'remote').
146 CONTEXT: processing remote data for replication origin "pg_16395" during "INSER
147 T" for replication target relation "public.test" in transaction 725 finished at
150 The LSN of the transaction that contains the change violating the
151 constraint and the replication origin name can be found from the server
152 log (LSN 0/14C0378 and replication origin pg_16395 in the above case).
153 The transaction that produced the conflict can be skipped by using
154 ALTER SUBSCRIPTION ... SKIP with the finish LSN (i.e., LSN 0/14C0378).
155 The finish LSN could be an LSN at which the transaction is committed or
156 prepared on the publisher. Alternatively, the transaction can also be
157 skipped by calling the pg_replication_origin_advance() function. Before
158 using this function, the subscription needs to be disabled temporarily
159 either by ALTER SUBSCRIPTION ... DISABLE or, the subscription can be
160 used with the disable_on_error option. Then, you can use
161 pg_replication_origin_advance() function with the node_name (i.e.,
162 pg_16395) and the next LSN of the finish LSN (i.e., 0/14C0379). The
163 current position of origins can be seen in the
164 pg_replication_origin_status system view. Please note that skipping the
165 whole transaction includes skipping changes that might not violate any
166 constraint. This can easily make the subscriber inconsistent. The
167 additional details regarding conflicting rows, such as their origin and
168 commit timestamp can be seen in the DETAIL line of the log. But note
169 that this information is only available when track_commit_timestamp is
170 enabled on the subscriber. Users can use this information to decide
171 whether to retain the local change or adopt the remote alteration. For
172 instance, the DETAIL line in the above log indicates that the existing
173 row was modified locally. Users can manually perform a
176 When the streaming mode is parallel, the finish LSN of failed
177 transactions may not be logged. In that case, it may be necessary to
178 change the streaming mode to on or off and cause the same conflicts
179 again so the finish LSN of the failed transaction will be written to
180 the server log. For the usage of finish LSN, please refer to ALTER
181 SUBSCRIPTION ... SKIP.