Merge pull request #266 from EngineeredVirus/main

[PG-159] pg_stat_monitor: Bucket start time should be aligned with fi…
yugabyte · Jun 21, 2022 · 2ba1960 · 2ba1960
2 parents d131040 + 7efc3fa
commit 2ba1960
Show file tree

Hide file tree

Showing 2 changed files with 47 additions and 55 deletions.
diff --git a/pg_stat_monitor.c b/pg_stat_monitor.c
@@ -2085,87 +2085,79 @@ static uint64
 get_next_wbucket(pgssSharedState *pgss)
 {
 	struct timeval	tv;
-	uint64			current_usec;
-	uint64			current_bucket_usec;
+	uint64			current_sec;
+	uint64			current_bucket_sec;
 	uint64			new_bucket_id;
 	uint64			prev_bucket_id;
 	struct tm		*lt;
-	bool			update_bucket = false;
+	char                    file_name[1024];
 
 	gettimeofday(&tv,NULL);
-	current_usec = (TimestampTz) tv.tv_sec - ((POSTGRES_EPOCH_JDATE - UNIX_EPOCH_JDATE) * SECS_PER_DAY);
-	current_usec = (current_usec * USECS_PER_SEC) + tv.tv_usec;
-	current_bucket_usec = pg_atomic_read_u64(&pgss->prev_bucket_usec);
+	current_sec = (TimestampTz) tv.tv_sec - ((POSTGRES_EPOCH_JDATE - UNIX_EPOCH_JDATE) * SECS_PER_DAY);
+	current_bucket_sec = pg_atomic_read_u64(&pgss->prev_bucket_sec);
 
 	/*
-	 * If current bucket expired we loop attempting to update prev_bucket_usec.
+	 * If current bucket expired we loop attempting to update prev_bucket_sec.
 	 *
 	 * pg_atomic_compare_exchange_u64 may fail in two possible ways:
 	 *    1. Another thread/process updated the variable before us.
 	 *    2. A spurious failure / hardware event.
 	 *
-	 * In both failure cases we read prev_bucket_usec from memory again, if it was
-	 * a spurious failure then the value of prev_bucket_usec must be the same as
+	 * In both failure cases we read prev_bucket_sec from memory again, if it was
+	 * a spurious failure then the value of prev_bucket_sec must be the same as
 	 * before, which will cause the while loop to execute again.
 	 *
-	 * If another thread updated prev_bucket_usec, then its current value will
+	 * If another thread updated prev_bucket_sec, then its current value will
 	 * definitely make the while condition to fail, we can stop the loop as another
-	 * thread has already updated prev_bucket_usec.
+	 * thread has already updated prev_bucket_sec.
 	 */
-	while ((current_usec - current_bucket_usec) > ((uint64)PGSM_BUCKET_TIME * 1000LU * 1000LU))
+	if ((current_sec - current_bucket_sec) < ((uint64)PGSM_BUCKET_TIME * 1000LU * 1000LU))
 	{
-		if (pg_atomic_compare_exchange_u64(&pgss->prev_bucket_usec, &current_bucket_usec, current_usec))
-		{
-			update_bucket = true;
-			break;
-		}
-
-		current_bucket_usec = pg_atomic_read_u64(&pgss->prev_bucket_usec);
+		return pg_atomic_read_u64(&pgss->current_wbucket);
 	}
 
-	if (update_bucket)
-	{
-		char          file_name[1024];
-		int            sec = 0;
-
-		new_bucket_id = (tv.tv_sec / PGSM_BUCKET_TIME) % PGSM_MAX_BUCKETS;
+	new_bucket_id = (tv.tv_sec / PGSM_BUCKET_TIME) % PGSM_MAX_BUCKETS;
 
-		/* Update bucket id and retrieve the previous one. */
-		prev_bucket_id = pg_atomic_exchange_u64(&pgss->current_wbucket, new_bucket_id);
+	/* Update bucket id and retrieve the previous one. */
+	prev_bucket_id = pg_atomic_exchange_u64(&pgss->current_wbucket, new_bucket_id);
 
-		LWLockAcquire(pgss->lock, LW_EXCLUSIVE);
-		hash_entry_dealloc(new_bucket_id, prev_bucket_id, pgss_qbuf);
+	tv.tv_sec = (tv.tv_sec) - (tv.tv_sec % PGSM_BUCKET_TIME);
+	lt = localtime(&tv.tv_sec);
 
-		if (pgss->overflow)
-		{
-			pgss->n_bucket_cycles += 1;
-			if (pgss->n_bucket_cycles >= PGSM_MAX_BUCKETS)
-			{
-				/*
-				 * A full rotation of PGSM_MAX_BUCKETS buckets happened since
-				 * we detected a query buffer overflow.
-				 * Reset overflow state and remove the dump file.
-				 */
-				pgss->overflow = false;
-				pgss->n_bucket_cycles = 0;
-				snprintf(file_name, 1024, "%s", PGSM_TEXT_FILE);
-				unlink(file_name);
-			}
-		}
+	LWLockAcquire(pgss->lock, LW_EXCLUSIVE);
 
+	/* Reconfirm that no other backend has created the bucket while we waited */
+	if (new_bucket_id == prev_bucket_id)
+	{
 		LWLockRelease(pgss->lock);
+		return new_bucket_id;
+	}
 
-		lt = localtime(&tv.tv_sec);
-		sec = lt->tm_sec - (lt->tm_sec % PGSM_BUCKET_TIME);
-		if (sec < 0)
-			sec = 0;
-		snprintf(pgss->bucket_start_time[new_bucket_id], sizeof(pgss->bucket_start_time[new_bucket_id]),
-				"%04d-%02d-%02d %02d:%02d:%02d", lt->tm_year + 1900, lt->tm_mon + 1, lt->tm_mday, lt->tm_hour, lt->tm_min, sec);
+	hash_entry_dealloc(new_bucket_id, prev_bucket_id, pgss_qbuf);
 
-		return new_bucket_id;
+	if (pgss->overflow)
+	{
+		pgss->n_bucket_cycles += 1;
+		if (pgss->n_bucket_cycles >= PGSM_MAX_BUCKETS)
+		{
+			/*
+				* A full rotation of PGSM_MAX_BUCKETS buckets happened since
+				* we detected a query buffer overflow.
+				* Reset overflow state and remove the dump file.
+				*/
+			pgss->overflow = false;
+			pgss->n_bucket_cycles = 0;
+			snprintf(file_name, 1024, "%s", PGSM_TEXT_FILE);
+			unlink(file_name);
+		}
 	}
 
-	return pg_atomic_read_u64(&pgss->current_wbucket);
+	snprintf(pgss->bucket_start_time[new_bucket_id], sizeof(pgss->bucket_start_time[new_bucket_id]),
+			"%04d-%02d-%02d %02d:%02d:%02d", lt->tm_year + 1900, lt->tm_mon + 1, lt->tm_mday, lt->tm_hour, lt->tm_min, lt->tm_sec);
+
+	LWLockRelease(pgss->lock);
+
+	return new_bucket_id;
 }
 
 #if PG_VERSION_NUM < 140000

diff --git a/pg_stat_monitor.h b/pg_stat_monitor.h
@@ -310,7 +310,7 @@ typedef struct pgssSharedState
 	Size				extent;				/* current extent of query file */
 	int64				n_writers;			/* number of active writers to query file */
 	pg_atomic_uint64	current_wbucket;
-	pg_atomic_uint64	prev_bucket_usec;
+	pg_atomic_uint64	prev_bucket_sec;
 	uint64				bucket_entry[MAX_BUCKETS];
 	char				bucket_start_time[MAX_BUCKETS][60];   	/* start time of the bucket */
 	LWLock				*errors_lock;		/* protects errors hashtable search/modification */
@@ -336,7 +336,7 @@ do { \
 		x->cur_median_usage = ASSUMED_MEDIAN_INIT; \
 		x->n_writers = 0; \
 		pg_atomic_init_u64(&x->current_wbucket, 0); \
-		pg_atomic_init_u64(&x->prev_bucket_usec, 0); \
+		pg_atomic_init_u64(&x->prev_bucket_sec, 0); \
 		memset(&x->bucket_entry, 0, MAX_BUCKETS * sizeof(uint64)); \
 } while(0)