#include "postgres.h"
#include <signal.h>
#include <sys/time.h>
#include <unistd.h>
#include "access/xlog.h"
#include "access/xlog_internal.h"
#include "libpq/pqsignal.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "postmaster/bgwriter.h"
#include "replication/syncrep.h"
#include "storage/bufmgr.h"
#include "storage/condition_variable.h"
#include "storage/fd.h"
#include "storage/ipc.h"
#include "storage/lwlock.h"
#include "storage/proc.h"
#include "storage/shmem.h"
#include "storage/smgr.h"
#include "storage/spin.h"
#include "utils/guc.h"
#include "utils/memutils.h"
#include "utils/resowner.h"

Include dependency graph for checkpointer.c:

Data Structures
struct	CheckpointerRequest

struct	CheckpointerShmemStruct

Macros
#define	WRITES_PER_ABSORB 1000

Functions
static void	CheckArchiveTimeout (void)

static bool	IsCheckpointOnSchedule (double progress)

static bool	ImmediateCheckpointRequested (void)

static bool	CompactCheckpointerRequestQueue (void)

static void	UpdateSharedMemoryConfig (void)

static void	chkpt_quickdie (SIGNAL_ARGS)

static void	ChkptSigHupHandler (SIGNAL_ARGS)

static void	ReqCheckpointHandler (SIGNAL_ARGS)

static void	chkpt_sigusr1_handler (SIGNAL_ARGS)

static void	ReqShutdownHandler (SIGNAL_ARGS)

void	CheckpointerMain (void)

void	CheckpointWriteDelay (int flags, double progress)

Size	CheckpointerShmemSize (void)

void	CheckpointerShmemInit (void)

void	RequestCheckpoint (int flags)

bool	ForwardFsyncRequest (RelFileNode rnode, ForkNumber forknum, BlockNumber segno)

void	AbsorbFsyncRequests (void)

bool	FirstCallSinceLastCheckpoint (void)

Variables
static CheckpointerShmemStruct *	CheckpointerShmem

int	CheckPointTimeout = 300

int	CheckPointWarning = 30

double	CheckPointCompletionTarget = 0.5

static volatile sig_atomic_t	got_SIGHUP = false

static volatile sig_atomic_t	checkpoint_requested = false

static volatile sig_atomic_t	shutdown_requested = false

static bool	ckpt_active = false

static pg_time_t	ckpt_start_time

static XLogRecPtr	ckpt_start_recptr

static double	ckpt_cached_elapsed

static pg_time_t	last_checkpoint_time

static pg_time_t	last_xlog_switch_time

Macro Definition Documentation

#define WRITES_PER_ABSORB 1000

Definition at line 140 of file checkpointer.c.

Referenced by CheckpointWriteDelay().

Function Documentation

void AbsorbFsyncRequests ( void )

Definition at line 1303 of file checkpointer.c.

References AmCheckpointerProcess, BgWriterStats, END_CRIT_SECTION, CheckpointerRequest::forknum, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), PgStat_MsgBgWriter::m_buf_fsync_backend, PgStat_MsgBgWriter::m_buf_written_backend, CheckpointerShmemStruct::num_backend_fsync, CheckpointerShmemStruct::num_backend_writes, CheckpointerShmemStruct::num_requests, palloc(), pfree(), RememberFsyncRequest(), CheckpointerShmemStruct::requests, CheckpointerRequest::rnode, CheckpointerRequest::segno, and START_CRIT_SECTION.

Referenced by CheckpointerMain(), CheckpointWriteDelay(), mdpostckpt(), and mdsync().

 {
     CheckpointerRequest *requests = NULL;
     CheckpointerRequest *request;
     int         n;
 
     if (!AmCheckpointerProcess())
         return;
 
     LWLockAcquire(CheckpointerCommLock, LW_EXCLUSIVE);
 
     /* Transfer stats counts into pending pgstats message */
     BgWriterStats.m_buf_written_backend += CheckpointerShmem->num_backend_writes;
     BgWriterStats.m_buf_fsync_backend += CheckpointerShmem->num_backend_fsync;
 
     CheckpointerShmem->num_backend_writes = 0;
     CheckpointerShmem->num_backend_fsync = 0;
 
     /*
      * We try to avoid holding the lock for a long time by copying the request
      * array, and processing the requests after releasing the lock.
      *
      * Once we have cleared the requests from shared memory, we have to PANIC
      * if we then fail to absorb them (eg, because our hashtable runs out of
      * memory).  This is because the system cannot run safely if we are unable
      * to fsync what we have been told to fsync.  Fortunately, the hashtable
      * is so small that the problem is quite unlikely to arise in practice.
      */
     n = CheckpointerShmem->num_requests;
     if (n > 0)
     {
         requests = (CheckpointerRequest *) palloc(n * sizeof(CheckpointerRequest));
         memcpy(requests, CheckpointerShmem->requests, n * sizeof(CheckpointerRequest));
     }
 
     START_CRIT_SECTION();
 
     CheckpointerShmem->num_requests = 0;
 
     LWLockRelease(CheckpointerCommLock);
 
     for (request = requests; n > 0; request++, n--)
         RememberFsyncRequest(request->rnode, request->forknum, request->segno);
 
     END_CRIT_SECTION();
 
     if (requests)
         pfree(requests);
 }

static void CheckArchiveTimeout ( void )

static

Definition at line 585 of file checkpointer.c.

References DEBUG1, elog, GetLastImportantRecPtr(), GetLastSegSwitchData(), last_xlog_switch_time, Max, now(), RecoveryInProgress(), RequestXLogSwitch(), wal_segment_size, XLogArchiveTimeout, and XLogSegmentOffset.

Referenced by CheckpointerMain(), and CheckpointWriteDelay().

 {
     pg_time_t   now;
     pg_time_t   last_time;
     XLogRecPtr  last_switch_lsn;
 
     if (XLogArchiveTimeout <= 0 || RecoveryInProgress())
         return;
 
     now = (pg_time_t) time(NULL);
 
     /* First we do a quick check using possibly-stale local state. */
     if ((int) (now - last_xlog_switch_time) < XLogArchiveTimeout)
         return;
 
     /*
      * Update local state ... note that last_xlog_switch_time is the last time
      * a switch was performed *or requested*.
      */
     last_time = GetLastSegSwitchData(&last_switch_lsn);
 
     last_xlog_switch_time = Max(last_xlog_switch_time, last_time);
 
     /* Now we can do the real checks */
     if ((int) (now - last_xlog_switch_time) >= XLogArchiveTimeout)
     {
         /*
          * Switch segment only when "important" WAL has been logged since the
          * last segment switch (last_switch_lsn points to end of segment
          * switch occurred in).
          */
         if (GetLastImportantRecPtr() > last_switch_lsn)
         {
             XLogRecPtr  switchpoint;
 
             /* mark switch as unimportant, avoids triggering checkpoints */
             switchpoint = RequestXLogSwitch(true);
 
             /*
              * If the returned pointer points exactly to a segment boundary,
              * assume nothing happened.
              */
             if (XLogSegmentOffset(switchpoint, wal_segment_size) != 0)
                 elog(DEBUG1, "write-ahead log switch forced (archive_timeout=%d)",
                      XLogArchiveTimeout);
         }
 
         /*
          * Update state in any case, so we don't retry constantly when the
          * system is idle.
          */
         last_xlog_switch_time = now;
     }
 }

void CheckpointerMain ( void )

Definition at line 193 of file checkpointer.c.

References AbortBufferIO(), AbsorbFsyncRequests(), ALLOCSET_DEFAULT_SIZES, AllocSetContextCreate(), AtEOXact_Buffers(), AtEOXact_Files(), AtEOXact_HashTables(), AtEOXact_SMgr(), BgWriterStats, BlockSig, CheckArchiveTimeout(), CHECKPOINT_CAUSE_TIME, CHECKPOINT_CAUSE_XLOG, CHECKPOINT_END_OF_RECOVERY, checkpoint_requested, CheckpointerShmemStruct::checkpointer_pid, PROC_HDR::checkpointerLatch, CheckPointTimeout, CheckPointWarning, chkpt_quickdie(), chkpt_sigusr1_handler(), ChkptSigHupHandler(), ckpt_active, ckpt_cached_elapsed, CheckpointerShmemStruct::ckpt_done, CheckpointerShmemStruct::ckpt_failed, CheckpointerShmemStruct::ckpt_flags, CheckpointerShmemStruct::ckpt_lck, ckpt_start_recptr, ckpt_start_time, CheckpointerShmemStruct::ckpt_started, ConditionVariableCancelSleep(), CreateCheckPoint(), CreateRestartPoint(), CurrentResourceOwner, EmitErrorReport(), ereport, errhint(), errmsg_plural(), error_context_stack, ExitOnAnyError, FlushErrorState(), GetInsertRecPtr(), GetXLogReplayRecPtr(), got_SIGHUP, HOLD_INTERRUPTS, last_checkpoint_time, last_xlog_switch_time, LOG, LWLockReleaseAll(), PgStat_MsgBgWriter::m_requested_checkpoints, PgStat_MsgBgWriter::m_timed_checkpoints, MemoryContextResetAndDeleteChildren, MemoryContextSwitchTo(), Min, MyLatch, MyProc, MyProcPid, now(), PG_exception_stack, PG_SETMASK, pg_usleep(), PGC_SIGHUP, pgstat_report_wait_end(), pgstat_send_bgwriter(), pqsignal(), proc_exit(), ProcessConfigFile(), ProcGlobal, PGPROC::procLatch, RecoveryInProgress(), ReqCheckpointHandler(), ReqShutdownHandler(), ResetLatch(), RESOURCE_RELEASE_BEFORE_LOCKS, ResourceOwnerCreate(), ResourceOwnerRelease(), RESUME_INTERRUPTS, shutdown_requested, ShutdownXLOG(), SIG_DFL, SIG_IGN, SIGALRM, SIGCHLD, SIGCONT, SIGHUP, SIGPIPE, SIGQUIT, SIGTTIN, SIGTTOU, SIGUSR1, SIGUSR2, SIGWINCH, smgrcloseall(), SpinLockAcquire, SpinLockRelease, TopMemoryContext, UnBlockSig, UnlockBuffers(), UpdateSharedMemoryConfig(), WAIT_EVENT_CHECKPOINTER_MAIN, WaitLatch(), WL_LATCH_SET, WL_POSTMASTER_DEATH, WL_TIMEOUT, and XLogArchiveTimeout.

Referenced by AuxiliaryProcessMain().

 {
     sigjmp_buf  local_sigjmp_buf;
     MemoryContext checkpointer_context;
 
     CheckpointerShmem->checkpointer_pid = MyProcPid;
 
     /*
      * Properly accept or ignore signals the postmaster might send us
      *
      * Note: we deliberately ignore SIGTERM, because during a standard Unix
      * system shutdown cycle, init will SIGTERM all processes at once.  We
      * want to wait for the backends to exit, whereupon the postmaster will
      * tell us it's okay to shut down (via SIGUSR2).
      */
     pqsignal(SIGHUP, ChkptSigHupHandler);   /* set flag to read config file */
     pqsignal(SIGINT, ReqCheckpointHandler); /* request checkpoint */
     pqsignal(SIGTERM, SIG_IGN); /* ignore SIGTERM */
     pqsignal(SIGQUIT, chkpt_quickdie);  /* hard crash time */
     pqsignal(SIGALRM, SIG_IGN);
     pqsignal(SIGPIPE, SIG_IGN);
     pqsignal(SIGUSR1, chkpt_sigusr1_handler);
     pqsignal(SIGUSR2, ReqShutdownHandler);  /* request shutdown */
 
     /*
      * Reset some signals that are accepted by postmaster but not here
      */
     pqsignal(SIGCHLD, SIG_DFL);
     pqsignal(SIGTTIN, SIG_DFL);
     pqsignal(SIGTTOU, SIG_DFL);
     pqsignal(SIGCONT, SIG_DFL);
     pqsignal(SIGWINCH, SIG_DFL);
 
     /* We allow SIGQUIT (quickdie) at all times */
     sigdelset(&BlockSig, SIGQUIT);
 
     /*
      * Initialize so that first time-driven event happens at the correct time.
      */
     last_checkpoint_time = last_xlog_switch_time = (pg_time_t) time(NULL);
 
     /*
      * Create a resource owner to keep track of our resources (currently only
      * buffer pins).
      */
     CurrentResourceOwner = ResourceOwnerCreate(NULL, "Checkpointer");
 
     /*
      * Create a memory context that we will do all our work in.  We do this so
      * that we can reset the context during error recovery and thereby avoid
      * possible memory leaks.  Formerly this code just ran in
      * TopMemoryContext, but resetting that would be a really bad idea.
      */
     checkpointer_context = AllocSetContextCreate(TopMemoryContext,
                                                  "Checkpointer",
                                                  ALLOCSET_DEFAULT_SIZES);
     MemoryContextSwitchTo(checkpointer_context);
 
     /*
      * If an exception is encountered, processing resumes here.
      *
      * See notes in postgres.c about the design of this coding.
      */
     if (sigsetjmp(local_sigjmp_buf, 1) != 0)
     {
         /* Since not using PG_TRY, must reset error stack by hand */
         error_context_stack = NULL;
 
         /* Prevent interrupts while cleaning up */
         HOLD_INTERRUPTS();
 
         /* Report the error to the server log */
         EmitErrorReport();
 
         /*
          * These operations are really just a minimal subset of
          * AbortTransaction().  We don't have very many resources to worry
          * about in checkpointer, but we do have LWLocks, buffers, and temp
          * files.
          */
         LWLockReleaseAll();
         ConditionVariableCancelSleep();
         pgstat_report_wait_end();
         AbortBufferIO();
         UnlockBuffers();
         /* buffer pins are released here: */
         ResourceOwnerRelease(CurrentResourceOwner,
                              RESOURCE_RELEASE_BEFORE_LOCKS,
                              false, true);
         /* we needn't bother with the other ResourceOwnerRelease phases */
         AtEOXact_Buffers(false);
         AtEOXact_SMgr();
         AtEOXact_Files();
         AtEOXact_HashTables(false);
 
         /* Warn any waiting backends that the checkpoint failed. */
         if (ckpt_active)
         {
             SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
             CheckpointerShmem->ckpt_failed++;
             CheckpointerShmem->ckpt_done = CheckpointerShmem->ckpt_started;
             SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
             ckpt_active = false;
         }
 
         /*
          * Now return to normal top-level context and clear ErrorContext for
          * next time.
          */
         MemoryContextSwitchTo(checkpointer_context);
         FlushErrorState();
 
         /* Flush any leaked data in the top-level context */
         MemoryContextResetAndDeleteChildren(checkpointer_context);
 
         /* Now we can allow interrupts again */
         RESUME_INTERRUPTS();
 
         /*
          * Sleep at least 1 second after any error.  A write error is likely
          * to be repeated, and we don't want to be filling the error logs as
          * fast as we can.
          */
         pg_usleep(1000000L);
 
         /*
          * Close all open files after any error.  This is helpful on Windows,
          * where holding deleted files open causes various strange errors.
          * It's not clear we need it elsewhere, but shouldn't hurt.
          */
         smgrcloseall();
     }
 
     /* We can now handle ereport(ERROR) */
     PG_exception_stack = &local_sigjmp_buf;
 
     /*
      * Unblock signals (they were blocked when the postmaster forked us)
      */
     PG_SETMASK(&UnBlockSig);
 
     /*
      * Ensure all shared memory values are set correctly for the config. Doing
      * this here ensures no race conditions from other concurrent updaters.
      */
     UpdateSharedMemoryConfig();
 
     /*
      * Advertise our latch that backends can use to wake us up while we're
      * sleeping.
      */
     ProcGlobal->checkpointerLatch = &MyProc->procLatch;
 
     /*
      * Loop forever
      */
     for (;;)
     {
         bool        do_checkpoint = false;
         int         flags = 0;
         pg_time_t   now;
         int         elapsed_secs;
         int         cur_timeout;
         int         rc;
 
         /* Clear any already-pending wakeups */
         ResetLatch(MyLatch);
 
         /*
          * Process any requests or signals received recently.
          */
         AbsorbFsyncRequests();
 
         if (got_SIGHUP)
         {
             got_SIGHUP = false;
             ProcessConfigFile(PGC_SIGHUP);
 
             /*
              * Checkpointer is the last process to shut down, so we ask it to
              * hold the keys for a range of other tasks required most of which
              * have nothing to do with checkpointing at all.
              *
              * For various reasons, some config values can change dynamically
              * so the primary copy of them is held in shared memory to make
              * sure all backends see the same value.  We make Checkpointer
              * responsible for updating the shared memory copy if the
              * parameter setting changes because of SIGHUP.
              */
             UpdateSharedMemoryConfig();
         }
         if (checkpoint_requested)
         {
             checkpoint_requested = false;
             do_checkpoint = true;
             BgWriterStats.m_requested_checkpoints++;
         }
         if (shutdown_requested)
         {
             /*
              * From here on, elog(ERROR) should end with exit(1), not send
              * control back to the sigsetjmp block above
              */
             ExitOnAnyError = true;
             /* Close down the database */
             ShutdownXLOG(0, 0);
             /* Normal exit from the checkpointer is here */
             proc_exit(0);       /* done */
         }
 
         /*
          * Force a checkpoint if too much time has elapsed since the last one.
          * Note that we count a timed checkpoint in stats only when this
          * occurs without an external request, but we set the CAUSE_TIME flag
          * bit even if there is also an external request.
          */
         now = (pg_time_t) time(NULL);
         elapsed_secs = now - last_checkpoint_time;
         if (elapsed_secs >= CheckPointTimeout)
         {
             if (!do_checkpoint)
                 BgWriterStats.m_timed_checkpoints++;
             do_checkpoint = true;
             flags |= CHECKPOINT_CAUSE_TIME;
         }
 
         /*
          * Do a checkpoint if requested.
          */
         if (do_checkpoint)
         {
             bool        ckpt_performed = false;
             bool        do_restartpoint;
 
             /*
              * Check if we should perform a checkpoint or a restartpoint. As a
              * side-effect, RecoveryInProgress() initializes TimeLineID if
              * it's not set yet.
              */
             do_restartpoint = RecoveryInProgress();
 
             /*
              * Atomically fetch the request flags to figure out what kind of a
              * checkpoint we should perform, and increase the started-counter
              * to acknowledge that we've started a new checkpoint.
              */
             SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
             flags |= CheckpointerShmem->ckpt_flags;
             CheckpointerShmem->ckpt_flags = 0;
             CheckpointerShmem->ckpt_started++;
             SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
             /*
              * The end-of-recovery checkpoint is a real checkpoint that's
              * performed while we're still in recovery.
              */
             if (flags & CHECKPOINT_END_OF_RECOVERY)
                 do_restartpoint = false;
 
             /*
              * We will warn if (a) too soon since last checkpoint (whatever
              * caused it) and (b) somebody set the CHECKPOINT_CAUSE_XLOG flag
              * since the last checkpoint start.  Note in particular that this
              * implementation will not generate warnings caused by
              * CheckPointTimeout < CheckPointWarning.
              */
             if (!do_restartpoint &&
                 (flags & CHECKPOINT_CAUSE_XLOG) &&
                 elapsed_secs < CheckPointWarning)
                 ereport(LOG,
                         (errmsg_plural("checkpoints are occurring too frequently (%d second apart)",
                                        "checkpoints are occurring too frequently (%d seconds apart)",
                                        elapsed_secs,
                                        elapsed_secs),
                          errhint("Consider increasing the configuration parameter \"max_wal_size\".")));
 
             /*
              * Initialize checkpointer-private variables used during
              * checkpoint.
              */
             ckpt_active = true;
             if (do_restartpoint)
                 ckpt_start_recptr = GetXLogReplayRecPtr(NULL);
             else
                 ckpt_start_recptr = GetInsertRecPtr();
             ckpt_start_time = now;
             ckpt_cached_elapsed = 0;
 
             /*
              * Do the checkpoint.
              */
             if (!do_restartpoint)
             {
                 CreateCheckPoint(flags);
                 ckpt_performed = true;
             }
             else
                 ckpt_performed = CreateRestartPoint(flags);
 
             /*
              * After any checkpoint, close all smgr files.  This is so we
              * won't hang onto smgr references to deleted files indefinitely.
              */
             smgrcloseall();
 
             /*
              * Indicate checkpoint completion to any waiting backends.
              */
             SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
             CheckpointerShmem->ckpt_done = CheckpointerShmem->ckpt_started;
             SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
             if (ckpt_performed)
             {
                 /*
                  * Note we record the checkpoint start time not end time as
                  * last_checkpoint_time.  This is so that time-driven
                  * checkpoints happen at a predictable spacing.
                  */
                 last_checkpoint_time = now;
             }
             else
             {
                 /*
                  * We were not able to perform the restartpoint (checkpoints
                  * throw an ERROR in case of error).  Most likely because we
                  * have not received any new checkpoint WAL records since the
                  * last restartpoint. Try again in 15 s.
                  */
                 last_checkpoint_time = now - CheckPointTimeout + 15;
             }
 
             ckpt_active = false;
         }
 
         /* Check for archive_timeout and switch xlog files if necessary. */
         CheckArchiveTimeout();
 
         /*
          * Send off activity statistics to the stats collector.  (The reason
          * why we re-use bgwriter-related code for this is that the bgwriter
          * and checkpointer used to be just one process.  It's probably not
          * worth the trouble to split the stats support into two independent
          * stats message types.)
          */
         pgstat_send_bgwriter();
 
         /*
          * Sleep until we are signaled or it's time for another checkpoint or
          * xlog file switch.
          */
         now = (pg_time_t) time(NULL);
         elapsed_secs = now - last_checkpoint_time;
         if (elapsed_secs >= CheckPointTimeout)
             continue;           /* no sleep for us ... */
         cur_timeout = CheckPointTimeout - elapsed_secs;
         if (XLogArchiveTimeout > 0 && !RecoveryInProgress())
         {
             elapsed_secs = now - last_xlog_switch_time;
             if (elapsed_secs >= XLogArchiveTimeout)
                 continue;       /* no sleep for us ... */
             cur_timeout = Min(cur_timeout, XLogArchiveTimeout - elapsed_secs);
         }
 
         rc = WaitLatch(MyLatch,
                        WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
                        cur_timeout * 1000L /* convert to ms */ ,
                        WAIT_EVENT_CHECKPOINTER_MAIN);
 
         /*
          * Emergency bailout if postmaster has died.  This is to avoid the
          * necessity for manual cleanup of all postmaster children.
          */
         if (rc & WL_POSTMASTER_DEATH)
             exit(1);
     }
 }

void CheckpointerShmemInit ( void )

Definition at line 926 of file checkpointer.c.

References CheckpointerShmemSize(), CheckpointerShmemStruct::ckpt_lck, CheckpointerShmemStruct::max_requests, MemSet, NBuffers, ShmemInitStruct(), and SpinLockInit.

Referenced by CreateSharedMemoryAndSemaphores().

 {
     Size        size = CheckpointerShmemSize();
     bool        found;
 
     CheckpointerShmem = (CheckpointerShmemStruct *)
         ShmemInitStruct("Checkpointer Data",
                         size,
                         &found);
 
     if (!found)
     {
         /*
          * First time through, so initialize.  Note that we zero the whole
          * requests array; this is so that CompactCheckpointerRequestQueue can
          * assume that any pad bytes in the request structs are zeroes.
          */
         MemSet(CheckpointerShmem, 0, size);
         SpinLockInit(&CheckpointerShmem->ckpt_lck);
         CheckpointerShmem->max_requests = NBuffers;
     }
 }

Size CheckpointerShmemSize ( void )

Definition at line 907 of file checkpointer.c.

References add_size(), mul_size(), NBuffers, and offsetof.

Referenced by CheckpointerShmemInit(), and CreateSharedMemoryAndSemaphores().

 {
     Size        size;
 
     /*
      * Currently, the size of the requests[] array is arbitrarily set equal to
      * NBuffers.  This may prove too large or small ...
      */
     size = offsetof(CheckpointerShmemStruct, requests);
     size = add_size(size, mul_size(NBuffers, sizeof(CheckpointerRequest)));
 
     return size;
 }

void CheckpointWriteDelay	(	int	flags,
		double	progress
	)

Definition at line 676 of file checkpointer.c.

References AbsorbFsyncRequests(), AmCheckpointerProcess, CheckArchiveTimeout(), CHECKPOINT_IMMEDIATE, got_SIGHUP, ImmediateCheckpointRequested(), IsCheckpointOnSchedule(), pg_usleep(), PGC_SIGHUP, pgstat_send_bgwriter(), ProcessConfigFile(), shutdown_requested, UpdateSharedMemoryConfig(), and WRITES_PER_ABSORB.

Referenced by BufferSync().

 {
     static int  absorb_counter = WRITES_PER_ABSORB;
 
     /* Do nothing if checkpoint is being executed by non-checkpointer process */
     if (!AmCheckpointerProcess())
         return;
 
     /*
      * Perform the usual duties and take a nap, unless we're behind schedule,
      * in which case we just try to catch up as quickly as possible.
      */
     if (!(flags & CHECKPOINT_IMMEDIATE) &&
         !shutdown_requested &&
         !ImmediateCheckpointRequested() &&
         IsCheckpointOnSchedule(progress))
     {
         if (got_SIGHUP)
         {
             got_SIGHUP = false;
             ProcessConfigFile(PGC_SIGHUP);
             /* update shmem copies of config variables */
             UpdateSharedMemoryConfig();
         }
 
         AbsorbFsyncRequests();
         absorb_counter = WRITES_PER_ABSORB;
 
         CheckArchiveTimeout();
 
         /*
          * Report interim activity statistics to the stats collector.
          */
         pgstat_send_bgwriter();
 
         /*
          * This sleep used to be connected to bgwriter_delay, typically 200ms.
          * That resulted in more frequent wakeups if not much work to do.
          * Checkpointer and bgwriter are no longer related so take the Big
          * Sleep.
          */
         pg_usleep(100000L);
     }
     else if (--absorb_counter <= 0)
     {
         /*
          * Absorb pending fsync requests after each WRITES_PER_ABSORB write
          * operations even when we don't sleep, to prevent overflow of the
          * fsync request queue.
          */
         AbsorbFsyncRequests();
         absorb_counter = WRITES_PER_ABSORB;
     }
 }

static void chkpt_quickdie ( SIGNAL_ARGS )

static

Definition at line 824 of file checkpointer.c.

References BlockSig, on_exit_reset(), and PG_SETMASK.

Referenced by CheckpointerMain().

 {
     PG_SETMASK(&BlockSig);
 
     /*
      * We DO NOT want to run proc_exit() callbacks -- we're here because
      * shared memory may be corrupted, so we don't want to try to clean up our
      * transaction.  Just nail the windows shut and get out of town.  Now that
      * there's an atexit callback to prevent third-party code from breaking
      * things by calling exit() directly, we have to reset the callbacks
      * explicitly to make this work as intended.
      */
     on_exit_reset();
 
     /*
      * Note we do exit(2) not exit(0).  This is to force the postmaster into a
      * system reset cycle if some idiot DBA sends a manual SIGQUIT to a random
      * backend.  This is necessary precisely because we don't clean up our
      * shared memory state.  (The "dead man switch" mechanism in pmsignal.c
      * should ensure the postmaster sees this as a crash, too, but no harm in
      * being doubly sure.)
      */
     exit(2);
 }

static void chkpt_sigusr1_handler ( SIGNAL_ARGS )

static

Definition at line 875 of file checkpointer.c.

References latch_sigusr1_handler().

Referenced by CheckpointerMain().

 {
     int         save_errno = errno;
 
     latch_sigusr1_handler();
 
     errno = save_errno;
 }

static void ChkptSigHupHandler ( SIGNAL_ARGS )

static

Definition at line 851 of file checkpointer.c.

References got_SIGHUP, MyLatch, and SetLatch().

Referenced by CheckpointerMain().

 {
     int         save_errno = errno;
 
     got_SIGHUP = true;
     SetLatch(MyLatch);
 
     errno = save_errno;
 }

static bool CompactCheckpointerRequestQueue ( void )

static

Definition at line 1195 of file checkpointer.c.

References Assert, CurrentMemoryContext, DEBUG1, HASHCTL::entrysize, ereport, errmsg(), HASH_BLOBS, HASH_CONTEXT, hash_create(), hash_destroy(), HASH_ELEM, HASH_ENTER, hash_search(), HASHCTL::hcxt, HASHCTL::keysize, LWLockHeldByMe(), MemSet, CheckpointerShmemStruct::num_requests, palloc0(), pfree(), and CheckpointerShmemStruct::requests.

Referenced by ForwardFsyncRequest().

 {
     struct CheckpointerSlotMapping
     {
         CheckpointerRequest request;
         int         slot;
     };
 
     int         n,
                 preserve_count;
     int         num_skipped = 0;
     HASHCTL     ctl;
     HTAB       *htab;
     bool       *skip_slot;
 
     /* must hold CheckpointerCommLock in exclusive mode */
     Assert(LWLockHeldByMe(CheckpointerCommLock));
 
     /* Initialize skip_slot array */
     skip_slot = palloc0(sizeof(bool) * CheckpointerShmem->num_requests);
 
     /* Initialize temporary hash table */
     MemSet(&ctl, 0, sizeof(ctl));
     ctl.keysize = sizeof(CheckpointerRequest);
     ctl.entrysize = sizeof(struct CheckpointerSlotMapping);
     ctl.hcxt = CurrentMemoryContext;
 
     htab = hash_create("CompactCheckpointerRequestQueue",
                        CheckpointerShmem->num_requests,
                        &ctl,
                        HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
 
     /*
      * The basic idea here is that a request can be skipped if it's followed
      * by a later, identical request.  It might seem more sensible to work
      * backwards from the end of the queue and check whether a request is
      * *preceded* by an earlier, identical request, in the hopes of doing less
      * copying.  But that might change the semantics, if there's an
      * intervening FORGET_RELATION_FSYNC or FORGET_DATABASE_FSYNC request, so
      * we do it this way.  It would be possible to be even smarter if we made
      * the code below understand the specific semantics of such requests (it
      * could blow away preceding entries that would end up being canceled
      * anyhow), but it's not clear that the extra complexity would buy us
      * anything.
      */
     for (n = 0; n < CheckpointerShmem->num_requests; n++)
     {
         CheckpointerRequest *request;
         struct CheckpointerSlotMapping *slotmap;
         bool        found;
 
         /*
          * We use the request struct directly as a hashtable key.  This
          * assumes that any padding bytes in the structs are consistently the
          * same, which should be okay because we zeroed them in
          * CheckpointerShmemInit.  Note also that RelFileNode had better
          * contain no pad bytes.
          */
         request = &CheckpointerShmem->requests[n];
         slotmap = hash_search(htab, request, HASH_ENTER, &found);
         if (found)
         {
             /* Duplicate, so mark the previous occurrence as skippable */
             skip_slot[slotmap->slot] = true;
             num_skipped++;
         }
         /* Remember slot containing latest occurrence of this request value */
         slotmap->slot = n;
     }
 
     /* Done with the hash table. */
     hash_destroy(htab);
 
     /* If no duplicates, we're out of luck. */
     if (!num_skipped)
     {
         pfree(skip_slot);
         return false;
     }
 
     /* We found some duplicates; remove them. */
     preserve_count = 0;
     for (n = 0; n < CheckpointerShmem->num_requests; n++)
     {
         if (skip_slot[n])
             continue;
         CheckpointerShmem->requests[preserve_count++] = CheckpointerShmem->requests[n];
     }
     ereport(DEBUG1,
             (errmsg("compacted fsync request queue from %d entries to %d entries",
                     CheckpointerShmem->num_requests, preserve_count)));
     CheckpointerShmem->num_requests = preserve_count;
 
     /* Cleanup. */
     pfree(skip_slot);
     return true;
 }

bool FirstCallSinceLastCheckpoint ( void )

Definition at line 1376 of file checkpointer.c.

References CheckpointerShmemStruct::ckpt_done, CheckpointerShmemStruct::ckpt_lck, SpinLockAcquire, and SpinLockRelease.

Referenced by BackgroundWriterMain().

 {
     static int  ckpt_done = 0;
     int         new_done;
     bool        FirstCall = false;
 
     SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
     new_done = CheckpointerShmem->ckpt_done;
     SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
     if (new_done != ckpt_done)
         FirstCall = true;
 
     ckpt_done = new_done;
 
     return FirstCall;
 }

bool ForwardFsyncRequest	(	RelFileNode	rnode,
		ForkNumber	forknum,
		BlockNumber	segno
	)

Definition at line 1123 of file checkpointer.c.

References AmBackgroundWriterProcess, AmCheckpointerProcess, CheckpointerShmemStruct::checkpointer_pid, PROC_HDR::checkpointerLatch, CompactCheckpointerRequestQueue(), elog, ERROR, CheckpointerRequest::forknum, IsUnderPostmaster, LW_EXCLUSIVE, LWLockAcquire(), LWLockRelease(), CheckpointerShmemStruct::max_requests, CheckpointerShmemStruct::num_backend_fsync, CheckpointerShmemStruct::num_backend_writes, CheckpointerShmemStruct::num_requests, ProcGlobal, CheckpointerShmemStruct::requests, CheckpointerRequest::rnode, CheckpointerRequest::segno, and SetLatch().

Referenced by ForgetDatabaseFsyncRequests(), ForgetRelationFsyncRequests(), register_dirty_segment(), and register_unlink().

 {
     CheckpointerRequest *request;
     bool        too_full;
 
     if (!IsUnderPostmaster)
         return false;           /* probably shouldn't even get here */
 
     if (AmCheckpointerProcess())
         elog(ERROR, "ForwardFsyncRequest must not be called in checkpointer");
 
     LWLockAcquire(CheckpointerCommLock, LW_EXCLUSIVE);
 
     /* Count all backend writes regardless of if they fit in the queue */
     if (!AmBackgroundWriterProcess())
         CheckpointerShmem->num_backend_writes++;
 
     /*
      * If the checkpointer isn't running or the request queue is full, the
      * backend will have to perform its own fsync request.  But before forcing
      * that to happen, we can try to compact the request queue.
      */
     if (CheckpointerShmem->checkpointer_pid == 0 ||
         (CheckpointerShmem->num_requests >= CheckpointerShmem->max_requests &&
          !CompactCheckpointerRequestQueue()))
     {
         /*
          * Count the subset of writes where backends have to do their own
          * fsync
          */
         if (!AmBackgroundWriterProcess())
             CheckpointerShmem->num_backend_fsync++;
         LWLockRelease(CheckpointerCommLock);
         return false;
     }
 
     /* OK, insert request */
     request = &CheckpointerShmem->requests[CheckpointerShmem->num_requests++];
     request->rnode = rnode;
     request->forknum = forknum;
     request->segno = segno;
 
     /* If queue is more than half full, nudge the checkpointer to empty it */
     too_full = (CheckpointerShmem->num_requests >=
                 CheckpointerShmem->max_requests / 2);
 
     LWLockRelease(CheckpointerCommLock);
 
     /* ... but not till after we release the lock */
     if (too_full && ProcGlobal->checkpointerLatch)
         SetLatch(ProcGlobal->checkpointerLatch);
 
     return true;
 }

static bool ImmediateCheckpointRequested ( void )

static

Definition at line 646 of file checkpointer.c.

References CHECKPOINT_IMMEDIATE, checkpoint_requested, CheckpointerShmem, and CheckpointerShmemStruct::ckpt_flags.

Referenced by CheckpointWriteDelay().

 {
     if (checkpoint_requested)
     {
         volatile CheckpointerShmemStruct *cps = CheckpointerShmem;
 
         /*
          * We don't need to acquire the ckpt_lck in this case because we're
          * only looking at a single flag bit.
          */
         if (cps->ckpt_flags & CHECKPOINT_IMMEDIATE)
             return true;
     }
     return false;
 }

static bool IsCheckpointOnSchedule ( double progress )

static

Definition at line 740 of file checkpointer.c.

References Assert, CheckPointCompletionTarget, CheckPointSegments, CheckPointTimeout, ckpt_active, ckpt_cached_elapsed, ckpt_start_recptr, ckpt_start_time, elapsed_time(), GetInsertRecPtr(), gettimeofday(), GetXLogReplayRecPtr(), RecoveryInProgress(), and wal_segment_size.

Referenced by CheckpointWriteDelay().

 {
     XLogRecPtr  recptr;
     struct timeval now;
     double      elapsed_xlogs,
                 elapsed_time;
 
     Assert(ckpt_active);
 
     /* Scale progress according to checkpoint_completion_target. */
     progress *= CheckPointCompletionTarget;
 
     /*
      * Check against the cached value first. Only do the more expensive
      * calculations once we reach the target previously calculated. Since
      * neither time or WAL insert pointer moves backwards, a freshly
      * calculated value can only be greater than or equal to the cached value.
      */
     if (progress < ckpt_cached_elapsed)
         return false;
 
     /*
      * Check progress against WAL segments written and CheckPointSegments.
      *
      * We compare the current WAL insert location against the location
      * computed before calling CreateCheckPoint. The code in XLogInsert that
      * actually triggers a checkpoint when CheckPointSegments is exceeded
      * compares against RedoRecptr, so this is not completely accurate.
      * However, it's good enough for our purposes, we're only calculating an
      * estimate anyway.
      *
      * During recovery, we compare last replayed WAL record's location with
      * the location computed before calling CreateRestartPoint. That maintains
      * the same pacing as we have during checkpoints in normal operation, but
      * we might exceed max_wal_size by a fair amount. That's because there can
      * be a large gap between a checkpoint's redo-pointer and the checkpoint
      * record itself, and we only start the restartpoint after we've seen the
      * checkpoint record. (The gap is typically up to CheckPointSegments *
      * checkpoint_completion_target where checkpoint_completion_target is the
      * value that was in effect when the WAL was generated).
      */
     if (RecoveryInProgress())
         recptr = GetXLogReplayRecPtr(NULL);
     else
         recptr = GetInsertRecPtr();
     elapsed_xlogs = (((double) (recptr - ckpt_start_recptr)) /
                      wal_segment_size) / CheckPointSegments;
 
     if (progress < elapsed_xlogs)
     {
         ckpt_cached_elapsed = elapsed_xlogs;
         return false;
     }
 
     /*
      * Check progress against time elapsed and checkpoint_timeout.
      */
     gettimeofday(&now, NULL);
     elapsed_time = ((double) ((pg_time_t) now.tv_sec - ckpt_start_time) +
                     now.tv_usec / 1000000.0) / CheckPointTimeout;
 
     if (progress < elapsed_time)
     {
         ckpt_cached_elapsed = elapsed_time;
         return false;
     }
 
     /* It looks like we're on schedule. */
     return true;
 }

static void ReqCheckpointHandler ( SIGNAL_ARGS )

static

Definition at line 863 of file checkpointer.c.

References checkpoint_requested, MyLatch, and SetLatch().

Referenced by CheckpointerMain().

 {
     int         save_errno = errno;
 
     checkpoint_requested = true;
     SetLatch(MyLatch);
 
     errno = save_errno;
 }

static void ReqShutdownHandler ( SIGNAL_ARGS )

static

Definition at line 886 of file checkpointer.c.

References MyLatch, SetLatch(), and shutdown_requested.

Referenced by CheckpointerMain().

 {
     int         save_errno = errno;
 
     shutdown_requested = true;
     SetLatch(MyLatch);
 
     errno = save_errno;
 }

void RequestCheckpoint ( int flags )

Definition at line 967 of file checkpointer.c.

References CHECK_FOR_INTERRUPTS, CHECKPOINT_IMMEDIATE, CHECKPOINT_WAIT, CheckpointerShmemStruct::checkpointer_pid, CheckpointerShmemStruct::ckpt_done, CheckpointerShmemStruct::ckpt_failed, CheckpointerShmemStruct::ckpt_flags, CheckpointerShmemStruct::ckpt_lck, CheckpointerShmemStruct::ckpt_started, CreateCheckPoint(), elog, ereport, errhint(), errmsg(), ERROR, IsPostmasterEnvironment, LOG, pg_usleep(), smgrcloseall(), SpinLockAcquire, and SpinLockRelease.

Referenced by createdb(), do_pg_start_backup(), dropdb(), DropTableSpace(), movedb(), standard_ProcessUtility(), StartupXLOG(), XLogPageRead(), and XLogWrite().

 {
     int         ntries;
     int         old_failed,
                 old_started;
 
     /*
      * If in a standalone backend, just do it ourselves.
      */
     if (!IsPostmasterEnvironment)
     {
         /*
          * There's no point in doing slow checkpoints in a standalone backend,
          * because there's no other backends the checkpoint could disrupt.
          */
         CreateCheckPoint(flags | CHECKPOINT_IMMEDIATE);
 
         /*
          * After any checkpoint, close all smgr files.  This is so we won't
          * hang onto smgr references to deleted files indefinitely.
          */
         smgrcloseall();
 
         return;
     }
 
     /*
      * Atomically set the request flags, and take a snapshot of the counters.
      * When we see ckpt_started > old_started, we know the flags we set here
      * have been seen by checkpointer.
      *
      * Note that we OR the flags with any existing flags, to avoid overriding
      * a "stronger" request by another backend.  The flag senses must be
      * chosen to make this work!
      */
     SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
 
     old_failed = CheckpointerShmem->ckpt_failed;
     old_started = CheckpointerShmem->ckpt_started;
     CheckpointerShmem->ckpt_flags |= flags;
 
     SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
     /*
      * Send signal to request checkpoint.  It's possible that the checkpointer
      * hasn't started yet, or is in process of restarting, so we will retry a
      * few times if needed.  Also, if not told to wait for the checkpoint to
      * occur, we consider failure to send the signal to be nonfatal and merely
      * LOG it.
      */
     for (ntries = 0;; ntries++)
     {
         if (CheckpointerShmem->checkpointer_pid == 0)
         {
             if (ntries >= 20)   /* max wait 2.0 sec */
             {
                 elog((flags & CHECKPOINT_WAIT) ? ERROR : LOG,
                      "could not request checkpoint because checkpointer not running");
                 break;
             }
         }
         else if (kill(CheckpointerShmem->checkpointer_pid, SIGINT) != 0)
         {
             if (ntries >= 20)   /* max wait 2.0 sec */
             {
                 elog((flags & CHECKPOINT_WAIT) ? ERROR : LOG,
                      "could not signal for checkpoint: %m");
                 break;
             }
         }
         else
             break;              /* signal sent successfully */
 
         CHECK_FOR_INTERRUPTS();
         pg_usleep(100000L);     /* wait 0.1 sec, then retry */
     }
 
     /*
      * If requested, wait for completion.  We detect completion according to
      * the algorithm given above.
      */
     if (flags & CHECKPOINT_WAIT)
     {
         int         new_started,
                     new_failed;
 
         /* Wait for a new checkpoint to start. */
         for (;;)
         {
             SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
             new_started = CheckpointerShmem->ckpt_started;
             SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
             if (new_started != old_started)
                 break;
 
             CHECK_FOR_INTERRUPTS();
             pg_usleep(100000L);
         }
 
         /*
          * We are waiting for ckpt_done >= new_started, in a modulo sense.
          */
         for (;;)
         {
             int         new_done;
 
             SpinLockAcquire(&CheckpointerShmem->ckpt_lck);
             new_done = CheckpointerShmem->ckpt_done;
             new_failed = CheckpointerShmem->ckpt_failed;
             SpinLockRelease(&CheckpointerShmem->ckpt_lck);
 
             if (new_done - new_started >= 0)
                 break;
 
             CHECK_FOR_INTERRUPTS();
             pg_usleep(100000L);
         }
 
         if (new_failed != old_failed)
             ereport(ERROR,
                     (errmsg("checkpoint request failed"),
                      errhint("Consult recent messages in the server log for details.")));
     }
 }

static void UpdateSharedMemoryConfig ( void )

static

Definition at line 1357 of file checkpointer.c.

References DEBUG2, elog, SyncRepUpdateSyncStandbysDefined(), and UpdateFullPageWrites().

Referenced by CheckpointerMain(), and CheckpointWriteDelay().

 {
     /* update global shmem state for sync rep */
     SyncRepUpdateSyncStandbysDefined();
 
     /*
      * If full_page_writes has been changed by SIGHUP, we update it in shared
      * memory and write an XLOG_FPW_CHANGE record.
      */
     UpdateFullPageWrites();
 
     elog(DEBUG2, "checkpointer updated shared memory configuration values");
 }