Streaming Replication Protocol
| PostgreSQL 9.3.24 Documentation | ||||
|---|---|---|---|---|
| Prev | Up | Chapter 48. Frontend/Backend Protocol | Next | |
To initiate streaming replication, the frontend sends the replication parameter in the startup message. This tells the backend to go into walsender mode, wherein a small set of replication commands can be issued instead of SQL statements. Only the simple query protocol can be used in walsender mode. The commands accepted in walsender mode are:
- IDENTIFY_SYSTEM
-
Requests the server to identify itself. Server replies with a result set of a single row, containing three fields:
- systemid
-
The unique system identifier identifying the cluster. This can be used to check that the base backup used to initialize the standby came from the same cluster.
- timeline
-
Current TimelineID. Also useful to check that the standby is consistent with the master.
- xlogpos
-
Current xlog flush location. Useful to get a known location in the transaction log where streaming can start.
- TIMELINE_HISTORY tli
-
Requests the server to send over the timeline history file for timeline tli . Server replies with a result set of a single row, containing two fields:
- filename
-
Filename of the timeline history file, e.g 00000002.history.
- content
-
Contents of the timeline history file.
- START_REPLICATION XXX/XXX [ TIMELINE tli ]
-
Instructs server to start streaming WAL, starting at WAL position XXX/XXX . If TIMELINE option is specified, streaming starts on timeline tli ; otherwise, the server's current timeline is selected. The server can reply with an error, e.g. if the requested section of WAL has already been recycled. On success, server responds with a CopyBothResponse message, and then starts to stream WAL to the frontend.
If the client requests a timeline that's not the latest, but is part of the history of the server, the server will stream all the WAL on that timeline starting from the requested startpoint, up to the point where the server switched to another timeline. If the client requests streaming at exactly the end of an old timeline, the server responds immediately with CommandComplete without entering COPY mode.
After streaming all the WAL on a timeline that is not the latest one, the server will end streaming by exiting the COPY mode. When the client acknowledges this by also exiting COPY mode, the server sends a result set with one row and two columns, indicating the next timeline in this server's history. The first column is the next timeline's ID, and the second column is the XLOG position where the switch happened. Usually, the switch position is the end of the WAL that was streamed, but there are corner cases where the server can send some WAL from the old timeline that it has not itself replayed before promoting. Finally, the server sends CommandComplete message, and is ready to accept a new command.
WAL data is sent as a series of CopyData messages. (This allows other information to be intermixed; in particular the server can send an ErrorResponse message if it encounters a failure after beginning to stream.) The payload of each CopyData message from server to the client contains a message of one of the following formats:
- XLogData (B)
-
- Byte1('w')
-
Identifies the message as WAL data.
- Int64
-
The starting point of the WAL data in this message.
- Int64
-
The current end of WAL on the server.
- Int64
-
The server's system clock at the time of transmission, as microseconds since midnight on 2000-01-01.
- Byte n
-
A section of the WAL data stream.
A single WAL record is never split across two XLogData messages. When a WAL record crosses a WAL page boundary, and is therefore already split using continuation records, it can be split at the page boundary. In other words, the first main WAL record and its continuation records can be sent in different XLogData messages.
- Primary keepalive message (B)
-
- Byte1('k')
-
Identifies the message as a sender keepalive.
- Int64
-
The current end of WAL on the server.
- Int64
-
The server's system clock at the time of transmission, as microseconds since midnight on 2000-01-01.
- Byte1
-
1 means that the client should reply to this message as soon as possible, to avoid a timeout disconnect. 0 otherwise.
The receiving process can send replies back to the sender at any time, using one of the following message formats (also in the payload of a CopyData message):
- Standby status update (F)
-
- Byte1('r')
-
Identifies the message as a receiver status update.
- Int64
-
The location of the last WAL byte + 1 received and written to disk in the standby.
- Int64
-
The location of the last WAL byte + 1 flushed to disk in the standby.
- Int64
-
The location of the last WAL byte + 1 applied in the standby.
- Int64
-
The client's system clock at the time of transmission, as microseconds since midnight on 2000-01-01.
- Byte1
-
If 1, the client requests the server to reply to this message immediately. This can be used to ping the server, to test if the connection is still healthy.
- Hot Standby feedback message (F)
-
- Byte1('h')
-
Identifies the message as a Hot Standby feedback message.
- Int64
-
The client's system clock at the time of transmission, as microseconds since midnight on 2000-01-01.
- Int32
-
The standby's current xmin. This may be 0, if the standby is sending notification that Hot Standby feedback will no longer be sent on this connection. Later non-zero messages may reinitiate the feedback mechanism.
- Int32
-
The standby's current epoch.
- BASE_BACKUP [ LABEL 'label' ] [ PROGRESS ] [ FAST ] [ WAL ] [ NOWAIT ]
-
Instructs the server to start streaming a base backup. The system will automatically be put in backup mode before the backup is started, and taken out of it when the backup is complete. The following options are accepted:
- LABEL 'label'
-
Sets the label of the backup. If none is specified, a backup label of base backup will be used. The quoting rules for the label are the same as a standard SQL string with standard_conforming_strings turned on.
- PROGRESS
-
Request information required to generate a progress report. This will send back an approximate size in the header of each tablespace, which can be used to calculate how far along the stream is done. This is calculated by enumerating all the file sizes once before the transfer is even started, and may as such have a negative impact on the performance - in particular it may take longer before the first data is streamed. Since the database files can change during the backup, the size is only approximate and may both grow and shrink between the time of approximation and the sending of the actual files.
- FAST
-
Request a fast checkpoint.
- WAL
-
Include the necessary WAL segments in the backup. This will include all the files between start and stop backup in the pg_xlog directory of the base directory tar file.
- NOWAIT
-
By default, the backup will wait until the last required xlog segment has been archived, or emit a warning if log archiving is not enabled. Specifying NOWAIT disables both the waiting and the warning, leaving the client responsible for ensuring the required log is available.
When the backup is started, the server will first send two ordinary result sets, followed by one or more CopyResponse results.
The first ordinary result set contains the starting position of the backup, in a single row with two columns. The first column contains the start position given in XLogRecPtr format, and the second column contains the corresponding timeline ID.
The second ordinary result set has one row for each tablespace. The fields in this row are:
- spcoid
-
The oid of the tablespace, or NULL if it's the base directory.
- spclocation
-
The full path of the tablespace directory, or NULL if it's the base directory.
- size
-
The approximate size of the tablespace, if progress report has been requested; otherwise it's NULL .
After the second regular result set, one or more CopyResponse results will be sent, one for PGDATA and one for each additional tablespace other than pg_default and pg_global . The data in the CopyResponse results will be a tar format (following the "ustar interchange format" specified in the POSIX 1003.1-2008 standard) dump of the tablespace contents, except that the two trailing blocks of zeroes specified in the standard are omitted. After the tar data is complete, a final ordinary result set will be sent, containing the WAL end position of the backup, in the same format as the start position.
The tar archive for the data directory and each tablespace will contain all files in the directories, regardless of whether they are PostgreSQL files or other files added to the same directory. The only excluded files are:
-
postmaster.pid
-
postmaster.opts
-
pg_xlog , including subdirectories. If the backup is run with WAL files included, a synthesized version of pg_xlog will be included, but it will only contain the files necessary for the backup to work, not the rest of the contents.
Owner, group and file mode are set if the underlying file system on the server supports it.
Once all tablespaces have been sent, a final regular result set will be sent. This result set contains the end position of the backup, given in XLogRecPtr format as a single column in a single row.