4 * PostgreSQL transaction log manager
6 * Portions Copyright (c) 1996-2010, PostgreSQL Global Development Group
7 * Portions Copyright (c) 1994, Regents of the University of California
9 * $PostgreSQL: pgsql/src/include/access/xlog.h,v 1.117 2010/09/15 10:35:05 heikki Exp $
14 #include "access/rmgr.h"
15 #include "access/xlogdefs.h"
16 #include "lib/stringinfo.h"
17 #include "storage/buf.h"
18 #include "utils/pg_crc.h"
19 #include "utils/timestamp.h"
23 * The overall layout of an XLOG record is:
24 * Fixed-size header (XLogRecord struct)
32 * where there can be zero to three backup blocks (as signaled by xl_info flag
33 * bits). XLogRecord structs always start on MAXALIGN boundaries in the WAL
34 * files, and we round up SizeOfXLogRecord so that the rmgr data is also
35 * guaranteed to begin on a MAXALIGN boundary. However, no padding is added
36 * to align BkpBlock structs or backup block data.
38 * NOTE: xl_len counts only the rmgr data, not the XLogRecord header,
39 * and also not any backup blocks. xl_tot_len counts everything. Neither
40 * length field is rounded up to an alignment boundary.
42 typedef struct XLogRecord
44 pg_crc32 xl_crc; /* CRC for this record */
45 XLogRecPtr xl_prev; /* ptr to previous record in log */
46 TransactionId xl_xid; /* xact id */
47 uint32 xl_tot_len; /* total len of entire record */
48 uint32 xl_len; /* total len of rmgr data */
49 uint8 xl_info; /* flag bits, see below */
50 RmgrId xl_rmid; /* resource manager for this record */
52 /* Depending on MAXALIGN, there are either 2 or 6 wasted bytes here */
54 /* ACTUAL LOG DATA FOLLOWS AT END OF STRUCT */
58 #define SizeOfXLogRecord MAXALIGN(sizeof(XLogRecord))
60 #define XLogRecGetData(record) ((char*) (record) + SizeOfXLogRecord)
63 * XLOG uses only low 4 bits of xl_info. High 4 bits may be used by rmgr.
65 #define XLR_INFO_MASK 0x0F
68 * If we backed up any disk blocks with the XLOG record, we use flag bits in
69 * xl_info to signal it. We support backup of up to 3 disk blocks per XLOG
72 #define XLR_BKP_BLOCK_MASK 0x0E /* all info bits used for bkp blocks */
73 #define XLR_MAX_BKP_BLOCKS 3
74 #define XLR_SET_BKP_BLOCK(iblk) (0x08 >> (iblk))
75 #define XLR_BKP_BLOCK_1 XLR_SET_BKP_BLOCK(0) /* 0x08 */
76 #define XLR_BKP_BLOCK_2 XLR_SET_BKP_BLOCK(1) /* 0x04 */
77 #define XLR_BKP_BLOCK_3 XLR_SET_BKP_BLOCK(2) /* 0x02 */
80 * Bit 0 of xl_info is set if the backed-up blocks could safely be removed
81 * from a compressed version of XLOG (that is, they are backed up only to
82 * prevent partial-page-write problems, and not to ensure consistency of PITR
83 * recovery). The compression algorithm would need to extract data from the
84 * blocks to create an equivalent non-full-page XLOG record.
86 #define XLR_BKP_REMOVABLE 0x01
89 #define SYNC_METHOD_FSYNC 0
90 #define SYNC_METHOD_FDATASYNC 1
91 #define SYNC_METHOD_OPEN 2 /* for O_SYNC */
92 #define SYNC_METHOD_FSYNC_WRITETHROUGH 3
93 #define SYNC_METHOD_OPEN_DSYNC 4 /* for O_DSYNC */
94 extern int sync_method;
97 * The rmgr data to be written by XLogInsert() is defined by a chain of
98 * one or more XLogRecData structs. (Multiple structs would be used when
99 * parts of the source data aren't physically adjacent in memory, or when
100 * multiple associated buffers need to be specified.)
102 * If buffer is valid then XLOG will check if buffer must be backed up
103 * (ie, whether this is first change of that page since last checkpoint).
104 * If so, the whole page contents are attached to the XLOG record, and XLOG
105 * sets XLR_BKP_BLOCK_X bit in xl_info. Note that the buffer must be pinned
106 * and exclusive-locked by the caller, so that it won't change under us.
107 * NB: when the buffer is backed up, we DO NOT insert the data pointed to by
108 * this XLogRecData struct into the XLOG record, since we assume it's present
109 * in the buffer. Therefore, rmgr redo routines MUST pay attention to
110 * XLR_BKP_BLOCK_X to know what is actually stored in the XLOG record.
111 * The i'th XLR_BKP_BLOCK bit corresponds to the i'th distinct buffer
112 * value (ignoring InvalidBuffer) appearing in the rdata chain.
114 * When buffer is valid, caller must set buffer_std to indicate whether the
115 * page uses standard pd_lower/pd_upper header fields. If this is true, then
116 * XLOG is allowed to omit the free space between pd_lower and pd_upper from
117 * the backed-up page image. Note that even when buffer_std is false, the
118 * page MUST have an LSN field as its first eight bytes!
120 * Note: data can be NULL to indicate no rmgr data associated with this chain
121 * entry. This can be sensible (ie, not a wasted entry) if buffer is valid.
122 * The implication is that the buffer has been changed by the operation being
123 * logged, and so may need to be backed up, but the change can be redone using
124 * only information already present elsewhere in the XLOG entry.
126 typedef struct XLogRecData
128 char *data; /* start of rmgr data to include */
129 uint32 len; /* length of rmgr data to include */
130 Buffer buffer; /* buffer associated with data, if any */
131 bool buffer_std; /* buffer has standard pd_lower/pd_upper */
132 struct XLogRecData *next; /* next struct in chain, or NULL */
135 extern PGDLLIMPORT TimeLineID ThisTimeLineID; /* current TLI */
138 * Prior to 8.4, all activity during recovery was carried out by the startup
139 * process. This local variable continues to be used in many parts of the
140 * code to indicate actions taken by RecoveryManagers. Other processes that
141 * potentially perform work during recovery should check RecoveryInProgress().
142 * See XLogCtl notes in xlog.c.
144 extern bool InRecovery;
147 * Like InRecovery, standbyState is only valid in the startup process.
148 * In all other processes it will have the value STANDBY_DISABLED (so
149 * InHotStandby will read as FALSE).
151 * In DISABLED state, we're performing crash recovery or hot standby was
152 * disabled in postgresql.conf.
154 * In INITIALIZED state, we've run InitRecoveryTransactionEnvironment, but
155 * we haven't yet processed a RUNNING_XACTS or shutdown-checkpoint WAL record
156 * to initialize our master-transaction tracking system.
158 * When the transaction tracking is initialized, we enter the SNAPSHOT_PENDING
159 * state. The tracked information might still be incomplete, so we can't allow
160 * connections yet, but redo functions must update the in-memory state when
163 * In SNAPSHOT_READY mode, we have full knowledge of transactions that are
164 * (or were) running in the master at the current WAL location. Snapshots
165 * can be taken, and read-only queries can be run.
171 STANDBY_SNAPSHOT_PENDING,
172 STANDBY_SNAPSHOT_READY
175 extern HotStandbyState standbyState;
177 #define InHotStandby (standbyState >= STANDBY_SNAPSHOT_PENDING)
180 * Recovery target type.
181 * Only set during a Point in Time recovery, not when standby_mode = on
185 RECOVERY_TARGET_UNSET,
188 } RecoveryTargetType;
190 extern XLogRecPtr XactLastRecEnd;
193 * Replication mode. This is used to identify how long transaction
194 * commit should wait for replication.
196 * REPLICATION_MODE_ASYNC doesn't make transaction commit wait for
197 * replication, i.e., asynchronous replication.
199 * REPLICATION_MODE_RECV makes transaction commit wait for XLOG
200 * records to be received on the standby.
202 * REPLICATION_MODE_FSYNC makes transaction commit wait for XLOG
203 * records to be received and fsync'd on the standby.
205 * REPLICATION_MODE_APPLY makes transaction commit wait for XLOG
206 * records to be received, fsync'd and applied on the standby.
208 typedef enum ReplicationMode
210 InvalidReplicationMode = -1,
211 REPLICATION_MODE_ASYNC = 0,
212 REPLICATION_MODE_RECV,
213 REPLICATION_MODE_FSYNC,
214 REPLICATION_MODE_APPLY
217 * NOTE: if you add a new mode, change MAXREPLICATIONMODE below
218 * and update the ReplicationModeNames array in xlog.c
222 #define MAXREPLICATIONMODE REPLICATION_MODE_APPLY
224 extern const char *ReplicationModeNames[];
225 extern ReplicationMode rplMode;
227 /* these variables are GUC parameters related to XLOG */
228 extern int CheckPointSegments;
229 extern int wal_keep_segments;
230 extern int XLOGbuffers;
231 extern int XLogArchiveTimeout;
232 extern bool XLogArchiveMode;
233 extern char *XLogArchiveCommand;
234 extern bool EnableHotStandby;
235 extern bool log_checkpoints;
238 typedef enum WalLevel
240 WAL_LEVEL_MINIMAL = 0,
242 WAL_LEVEL_HOT_STANDBY
244 extern int wal_level;
246 #define XLogArchivingActive() (XLogArchiveMode && wal_level >= WAL_LEVEL_ARCHIVE)
247 #define XLogArchiveCommandSet() (XLogArchiveCommand[0] != '\0')
250 * Is WAL-logging necessary for archival or log-shipping, or can we skip
251 * WAL-logging if we fsync() the data before committing instead?
253 #define XLogIsNeeded() (wal_level >= WAL_LEVEL_ARCHIVE)
255 /* Do we need to WAL-log information required only for Hot Standby? */
256 #define XLogStandbyInfoActive() (wal_level >= WAL_LEVEL_HOT_STANDBY)
259 extern bool XLOG_DEBUG;
263 * OR-able request flag bits for checkpoints. The "cause" bits are used only
264 * for logging purposes. Note: the flags must be defined so that it's
265 * sensible to OR together request flags arising from different requestors.
268 /* These directly affect the behavior of CreateCheckPoint and subsidiaries */
269 #define CHECKPOINT_IS_SHUTDOWN 0x0001 /* Checkpoint is for shutdown */
270 #define CHECKPOINT_END_OF_RECOVERY 0x0002 /* Like shutdown checkpoint,
271 * but issued at end of WAL
273 #define CHECKPOINT_IMMEDIATE 0x0004 /* Do it without delays */
274 #define CHECKPOINT_FORCE 0x0008 /* Force even if no activity */
275 /* These are important to RequestCheckpoint */
276 #define CHECKPOINT_WAIT 0x0010 /* Wait for completion */
277 /* These indicate the cause of a checkpoint request */
278 #define CHECKPOINT_CAUSE_XLOG 0x0020 /* XLOG consumption */
279 #define CHECKPOINT_CAUSE_TIME 0x0040 /* Elapsed time */
281 /* Checkpoint statistics */
282 typedef struct CheckpointStatsData
284 TimestampTz ckpt_start_t; /* start of checkpoint */
285 TimestampTz ckpt_write_t; /* start of flushing buffers */
286 TimestampTz ckpt_sync_t; /* start of fsyncs */
287 TimestampTz ckpt_sync_end_t; /* end of fsyncs */
288 TimestampTz ckpt_end_t; /* end of checkpoint */
290 int ckpt_bufs_written; /* # of buffers written */
292 int ckpt_segs_added; /* # of new xlog segments created */
293 int ckpt_segs_removed; /* # of xlog segments deleted */
294 int ckpt_segs_recycled; /* # of xlog segments recycled */
295 } CheckpointStatsData;
297 extern CheckpointStatsData CheckpointStats;
299 extern XLogRecPtr XLogInsert(RmgrId rmid, uint8 info, XLogRecData *rdata);
300 extern void XLogFlush(XLogRecPtr RecPtr);
301 extern void XLogBackgroundFlush(void);
302 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
303 extern int XLogFileInit(uint32 log, uint32 seg,
304 bool *use_existent, bool use_lock);
305 extern int XLogFileOpen(uint32 log, uint32 seg);
308 extern void XLogGetLastRemoved(uint32 *log, uint32 *seg);
309 extern void XLogSetAsyncXactLSN(XLogRecPtr record);
311 extern void RestoreBkpBlocks(XLogRecPtr lsn, XLogRecord *record, bool cleanup);
313 extern void xlog_redo(XLogRecPtr lsn, XLogRecord *record);
314 extern void xlog_desc(StringInfo buf, uint8 xl_info, char *rec);
316 extern void issue_xlog_fsync(int fd, uint32 log, uint32 seg);
318 extern bool RecoveryInProgress(void);
319 extern bool XLogInsertAllowed(void);
320 extern void GetXLogReceiptTime(TimestampTz *rtime, bool *fromStream);
322 extern void UpdateControlFile(void);
323 extern uint64 GetSystemIdentifier(void);
324 extern Size XLOGShmemSize(void);
325 extern void XLOGShmemInit(void);
326 extern void BootStrapXLOG(void);
327 extern void StartupXLOG(void);
328 extern void ShutdownXLOG(int code, Datum arg);
329 extern void InitXLOGAccess(void);
330 extern void CreateCheckPoint(int flags);
331 extern bool CreateRestartPoint(int flags);
332 extern void XLogPutNextOid(Oid nextOid);
333 extern XLogRecPtr GetRedoRecPtr(void);
334 extern XLogRecPtr GetInsertRecPtr(void);
335 extern XLogRecPtr GetFlushRecPtr(void);
336 extern void GetNextXidAndEpoch(TransactionId *xid, uint32 *epoch);
337 extern TimeLineID GetRecoveryTargetTLI(void);
339 extern void HandleStartupProcInterrupts(void);
340 extern void StartupProcessMain(void);
341 extern void WakeupRecovery(void);
343 extern ReplicationMode ReplicationModeNameGetValue(char *name);