From 20b70d9d49ab785299df156e1c424a506a4ba2f0 Mon Sep 17 00:00:00 2001
From: Andres Freund <andres@anarazel.de>
Date: Wed, 15 Jul 2020 15:35:07 -0700
Subject: [PATCH v12 7/7] snapshot scalability: cache snapshots using a xact
 completion counter.

Previous commits made it faster/more scalable to compute snapshots. But not
building a snapshot is still faster. Now that GetSnapshotData() does not
maintain RecentGlobal* anymore, that is actually not too hard:

This commit introduces xactCompletionCount, which tracks the number of
top-level transactions with xids (i.e. which may have modified the database)
that completed in some form since the start of the server.

We can avoid rebuilding the snapshot's contents whenever the current
xactCompletionCount is the same as it was when the snapshot was
originally built.  Currently this check happens while holding
ProcArrayLock. While it's likely possible to perform the check before
acquiring ProcArrayLock, it's too complicated for now.

Author: Andres Freund
Reviewed-By: Robert Haas, Thomas Munro, David Rowley
Discussion: https://postgr.es/m/20200301083601.ews6hz5dduc3w2se@alap3.anarazel.de
---
 src/include/access/transam.h                |   9 ++
 src/include/utils/snapshot.h                |   7 ++
 src/backend/replication/logical/snapbuild.c |   1 +
 src/backend/storage/ipc/procarray.c         | 125 ++++++++++++++++----
 src/backend/utils/time/snapmgr.c            |   4 +
 5 files changed, 126 insertions(+), 20 deletions(-)

diff --git a/src/include/access/transam.h b/src/include/access/transam.h
index cf221582cd0..12fb7487b93 100644
--- a/src/include/access/transam.h
+++ b/src/include/access/transam.h
@@ -231,6 +231,15 @@ typedef struct VariableCacheData
 	FullTransactionId latestCompletedFullXid;	/* newest full XID that has
 												 * committed or aborted */
 
+	/*
+	 * Number of top-level transactions with xids (i.e. which may have
+	 * modified the database) that completed in some form since the start of
+	 * the server. This currently is solely used to check whether
+	 * GetSnapshotData() needs to recompute the contents of the snapshot, or
+	 * not. There are likely other users of this.  Always above 1.
+	 */
+	uint64 xactCompletionCount;
+
 	/*
 	 * These fields are protected by XactTruncationLock
 	 */
diff --git a/src/include/utils/snapshot.h b/src/include/utils/snapshot.h
index 35b1f05bea6..dea072e5edf 100644
--- a/src/include/utils/snapshot.h
+++ b/src/include/utils/snapshot.h
@@ -207,6 +207,13 @@ typedef struct SnapshotData
 
 	TimestampTz whenTaken;		/* timestamp when snapshot was taken */
 	XLogRecPtr	lsn;			/* position in the WAL stream when taken */
+
+	/*
+	 * The transaction completion count at the time GetSnapshotData() built
+	 * this snapshot. Allows to avoid re-computing static snapshots when no
+	 * transactions completed since the last GetSnapshotData().
+	 */
+	uint64		snapXactCompletionCount;
 } SnapshotData;
 
 #endif							/* SNAPSHOT_H */
diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index e9701ea7221..9d5d68f3fa7 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -524,6 +524,7 @@ SnapBuildBuildSnapshot(SnapBuild *builder)
 	snapshot->curcid = FirstCommandId;
 	snapshot->active_count = 0;
 	snapshot->regd_count = 0;
+	snapshot->snapXactCompletionCount = 0;
 
 	return snapshot;
 }
diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c
index 4f6da7c86a3..5078225c1ec 100644
--- a/src/backend/storage/ipc/procarray.c
+++ b/src/backend/storage/ipc/procarray.c
@@ -406,6 +406,7 @@ CreateSharedProcArray(void)
 		procArray->lastOverflowedXid = InvalidTransactionId;
 		procArray->replication_slot_xmin = InvalidTransactionId;
 		procArray->replication_slot_catalog_xmin = InvalidTransactionId;
+		ShmemVariableCache->xactCompletionCount = 1;
 	}
 
 	allProcs = ProcGlobal->allProcs;
@@ -533,6 +534,9 @@ ProcArrayRemove(PGPROC *proc, TransactionId latestXid)
 		/* Advance global latestCompletedXid while holding the lock */
 		MaintainLatestCompletedXid(latestXid);
 
+		/* Same with xactCompletionCount  */
+		ShmemVariableCache->xactCompletionCount++;
+
 		ProcGlobal->xids[proc->pgxactoff] = 0;
 		ProcGlobal->subxidStates[proc->pgxactoff].overflowed = false;
 		ProcGlobal->subxidStates[proc->pgxactoff].count = 0;
@@ -667,6 +671,7 @@ ProcArrayEndTransactionInternal(PGPROC *proc, TransactionId latestXid)
 {
 	size_t		pgxactoff = proc->pgxactoff;
 
+	Assert(LWLockHeldByMe(ProcArrayLock));
 	Assert(TransactionIdIsValid(ProcGlobal->xids[pgxactoff]));
 	Assert(ProcGlobal->xids[pgxactoff] == proc->xid);
 
@@ -698,6 +703,9 @@ ProcArrayEndTransactionInternal(PGPROC *proc, TransactionId latestXid)
 
 	/* Also advance global latestCompletedXid while holding the lock */
 	MaintainLatestCompletedXid(latestXid);
+
+	/* Same with xactCompletionCount  */
+	ShmemVariableCache->xactCompletionCount++;
 }
 
 /*
@@ -1910,6 +1918,93 @@ GetMaxSnapshotSubxidCount(void)
 	return TOTAL_MAX_CACHED_SUBXIDS;
 }
 
+/*
+ * Initialize old_snapshot_threshold specific parts of a newly build snapshot.
+ */
+static void
+GetSnapshotDataInitOldSnapshot(Snapshot snapshot)
+{
+	if (!OldSnapshotThresholdActive())
+	{
+		/*
+		 * If not using "snapshot too old" feature, fill related fields with
+		 * dummy values that don't require any locking.
+		 */
+		snapshot->lsn = InvalidXLogRecPtr;
+		snapshot->whenTaken = 0;
+	}
+	else
+	{
+		/*
+		 * Capture the current time and WAL stream location in case this
+		 * snapshot becomes old enough to need to fall back on the special
+		 * "old snapshot" logic.
+		 */
+		snapshot->lsn = GetXLogInsertRecPtr();
+		snapshot->whenTaken = GetSnapshotCurrentTimestamp();
+		MaintainOldSnapshotTimeMapping(snapshot->whenTaken, snapshot->xmin);
+	}
+}
+
+/*
+ * Helper function for GetSnapshotData() that check if the bulk of the
+ * visibility information in the snapshot is still valid. If so, it updates
+ * the fields that need to change and returns true. Otherwise it returns
+ * false.
+ *
+ * This very likely can be evolved to not need ProcArrayLock held (at very
+ * least in the case we already hold a snapshot), but that's for another day.
+ */
+static bool
+GetSnapshotDataReuse(Snapshot snapshot)
+{
+	uint64 curXactCompletionCount;
+
+	Assert(LWLockHeldByMe(ProcArrayLock));
+
+	if (unlikely(snapshot->snapXactCompletionCount == 0))
+		return false;
+
+	curXactCompletionCount = ShmemVariableCache->xactCompletionCount;
+	if (curXactCompletionCount != snapshot->snapXactCompletionCount)
+		return false;
+
+	/*
+	 * If the current xactCompletionCount is still the same as it was at the
+	 * time the snapshot was built, we can be sure that rebuilding the
+	 * contents of the snapshot the hard way would result in the same snapshot
+	 * contents:
+	 *
+	 * As explained in transam/README, the set of xids considered running by
+	 * GetSnapshotData() cannot change while ProcArrayLock is held. Snapshot
+	 * contents only depend on transactions with xids and xactCompletionCount
+	 * is incremented whenever a transaction with an xid finishes (while
+	 * holding ProcArrayLock) exclusively). Thus the xactCompletionCount check
+	 * ensures we would detect if the snapshot would have changed.
+	 *
+	 * As the snapshot contents are the same as it was before, it is is safe
+	 * to re-enter the snapshot's xmin into the PGPROC array. None of the rows
+	 * visible under the snapshot could already have been removed (that'd
+	 * require the set of running transactions to change) and it fulfills the
+	 * requirement that concurrent GetSnapshotData() calls yield the same
+	 * xmin.
+	 */
+	if (!TransactionIdIsValid(MyProc->xmin))
+		MyProc->xmin = TransactionXmin = snapshot->xmin;
+
+	RecentXmin = snapshot->xmin;
+	Assert(TransactionIdPrecedesOrEquals(TransactionXmin, RecentXmin));
+
+	snapshot->curcid = GetCurrentCommandId(false);
+	snapshot->active_count = 0;
+	snapshot->regd_count = 0;
+	snapshot->copied = false;
+
+	GetSnapshotDataInitOldSnapshot(snapshot);
+
+	return true;
+}
+
 /*
  * GetSnapshotData -- returns information about running transactions.
  *
@@ -1958,6 +2053,7 @@ GetSnapshotData(Snapshot snapshot)
 	TransactionId oldestxid;
 	int			mypgxactoff;
 	TransactionId myxid;
+	uint64		curXactCompletionCount;
 
 	TransactionId replication_slot_xmin = InvalidTransactionId;
 	TransactionId replication_slot_catalog_xmin = InvalidTransactionId;
@@ -2002,12 +2098,19 @@ GetSnapshotData(Snapshot snapshot)
 	 */
 	LWLockAcquire(ProcArrayLock, LW_SHARED);
 
+	if (GetSnapshotDataReuse(snapshot))
+	{
+		LWLockRelease(ProcArrayLock);
+		return snapshot;
+	}
+
 	latest_completed = ShmemVariableCache->latestCompletedFullXid;
 	mypgxactoff = MyProc->pgxactoff;
 	myxid = other_xids[mypgxactoff];
 	Assert(myxid == MyProc->xid);
 
 	oldestxid = ShmemVariableCache->oldestXid;
+	curXactCompletionCount = ShmemVariableCache->xactCompletionCount;
 
 	/* xmax is always latestCompletedFullXid + 1 */
 	xmax = XidFromFullTransactionId(latest_completed);
@@ -2261,6 +2364,7 @@ GetSnapshotData(Snapshot snapshot)
 	snapshot->xcnt = count;
 	snapshot->subxcnt = subcount;
 	snapshot->suboverflowed = suboverflowed;
+	snapshot->snapXactCompletionCount = curXactCompletionCount;
 
 	snapshot->curcid = GetCurrentCommandId(false);
 
@@ -2272,26 +2376,7 @@ GetSnapshotData(Snapshot snapshot)
 	snapshot->regd_count = 0;
 	snapshot->copied = false;
 
-	if (old_snapshot_threshold < 0)
-	{
-		/*
-		 * If not using "snapshot too old" feature, fill related fields with
-		 * dummy values that don't require any locking.
-		 */
-		snapshot->lsn = InvalidXLogRecPtr;
-		snapshot->whenTaken = 0;
-	}
-	else
-	{
-		/*
-		 * Capture the current time and WAL stream location in case this
-		 * snapshot becomes old enough to need to fall back on the special
-		 * "old snapshot" logic.
-		 */
-		snapshot->lsn = GetXLogInsertRecPtr();
-		snapshot->whenTaken = GetSnapshotCurrentTimestamp();
-		MaintainOldSnapshotTimeMapping(snapshot->whenTaken, xmin);
-	}
+	GetSnapshotDataInitOldSnapshot(snapshot);
 
 	return snapshot;
 }
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 689a3b6a597..09ea03c2063 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -595,6 +595,8 @@ SetTransactionSnapshot(Snapshot sourcesnap, VirtualTransactionId *sourcevxid,
 	CurrentSnapshot->takenDuringRecovery = sourcesnap->takenDuringRecovery;
 	/* NB: curcid should NOT be copied, it's a local matter */
 
+	CurrentSnapshot->snapXactCompletionCount = 0;
+
 	/*
 	 * Now we have to fix what GetSnapshotData did with MyProc->xmin and
 	 * TransactionXmin.  There is a race condition: to make sure we are not
@@ -670,6 +672,7 @@ CopySnapshot(Snapshot snapshot)
 	newsnap->regd_count = 0;
 	newsnap->active_count = 0;
 	newsnap->copied = true;
+	newsnap->snapXactCompletionCount = 0;
 
 	/* setup XID array */
 	if (snapshot->xcnt > 0)
@@ -2207,6 +2210,7 @@ RestoreSnapshot(char *start_address)
 	snapshot->curcid = serialized_snapshot.curcid;
 	snapshot->whenTaken = serialized_snapshot.whenTaken;
 	snapshot->lsn = serialized_snapshot.lsn;
+	snapshot->snapXactCompletionCount = 0;
 
 	/* Copy XIDs, if present. */
 	if (serialized_snapshot.xcnt > 0)
-- 
2.25.0.114.g5b0ca878e0

