Using io_method=sync in postgresql.conf file is causing psql client to fail with `database "postgres" does not exist`

[VikasPatnala]

When using io_method=sync in postgresql.conf file and start the client as follows is causing client to fail with error psql: error: connection to server on socket "/tmp/.s.PGSQL.5432" failed: FATAL: database "postgres" does not exist here is the demonstration.

1. Initializing the database
   • $ bin/initdb -D mydb
2. Change the io_method parameter to sync
   • $ vim mydb/postgresql.conf
   • change io_method to sync
3. Start the postgres server
   • $ bin/pg_ctl -D mydb -l logfile start -o "-d 5"
4. Start the client
   • $ bin/psql -d postgres
     psql: error: connection to server on socket "/tmp/.s.PGSQL.5432" failed: FATAL: database "postgres" does not exist

Here postgres, template0, template1 are the default databases that postgresql will create in the initdb phase and these are created as intended that we checked with invoking postgres in single user mode. (references for this in documentation: https://www.postgresql.org/docs/current/creating-cluster.html#CREATING-CLUSTER)

[VikasPatnala]

The message "database postgres does not exist" is coming from the function InitPostgres in file src/backend/utils/init/postinit.c, below is the code for that

else if (in_dbname != NULL)
  	{
                HeapTuple       tuple;
                Form_pg_database dbform;
 
  		// 29Sep - Vikas debug start
                // tuple = GetDatabaseTuple(in_dbname);
                tuple = GetDatabaseTupleByOid(Template1DbOid);
  		// 29Sep - Vikas debug end
  		if (!HeapTupleIsValid(tuple))
                        ereport(FATAL,
   					(errcode(ERRCODE_UNDEFINED_DATABASE),
                                         errmsg("database \"%s\" does not exist", in_dbname)));
		dbform = (Form_pg_database) GETSTRUCT(tuple);
                dboid = dbform->oid;
        }

The tuple returned by the function GetDatabaseTuple is not a valid tuple.
The function GetDatabaseTuple calls the function systable_getnext which is causing the problem.

In a successful case this function return a valid tuple whereas in failure this function returns invalid tuple.

static HeapTuple
GetDatabaseTuple(const char *dbname)
{
 	HeapTuple	tuple;
	Relation	relation;
	SysScanDesc scan;
	ScanKeyData key[1];

	/*
         * form a scan key
         */
	ScanKeyInit(&key[0],
                            	Anum_pg_database_datname,
                                BTEqualStrategyNumber, F_NAMEEQ,
                                CStringGetDatum(dbname));

	/*
         * Open pg_database and fetch a tuple.  Force heap scan if we haven't yet
         * built the critical shared relcache entries (i.e., we're starting up
         * without a shared relcache cache file).
         */
	relation = table_open(DatabaseRelationId, AccessShareLock);
	scan = systable_beginscan(relation, DatabaseNameIndexId,
                                                          criticalSharedRelcachesBuilt,
                                                          NULL,
                                                          1, key);

problem here =>	tuple = systable_getnext(scan); 

	/* Must copy tuple before releasing buffer */
	if (HeapTupleIsValid(tuple))
		tuple = heap_copytuple(tuple);

	/* all done */
	systable_endscan(scan);
	table_close(relation, AccessShareLock);

	return tuple;
}