>Tab1
>Key1   Value1
>1          1000
>2          2000
>3          3000
>4          4000
> 
>Tab2
>Key2       Key1       Value2
>1             1               400
>2             1               400
>3             2             2000
>4             3             1000
>5             3             1000
>6             3               500
> 
>Result
>Key1       Key2       Value
>1             1               400
>1             2               400
>1             null            200
>2             3             2000
>3             4             1000
>3             5             1000
>3             6               500
>3             null            500
>4             null          4000
>
>Note:
>In Tab2 are 0 to n related records for each record in Tab1.
>In return dataset should be one record with key2 is null if value1 <> sum(value2) for one relation.

Assuming Key1 and Key2 are primary keys, this should be quite simple:

WITH TabSum (Key1, Value1, Value2) as
(SELECT t1.Key1, t1.Value1, sum(t2.Value2)
FROM Tab1 t1
LEFT JOIN Tab2 t2 ON t1.Key1 = t2.Key1
GROUP BY 1, 2)

SELECT Key1, Key2, Value2
FROM Tab2
UNION
SELECT Key1, null, Value1 - coalesce(Value2, 0)
FROM TabSum
WHERE Value1 is distinct from Value2

I'm uncertain whether you need to add ORDER BY 1, 2 NULLS LAST to this statement or not.

HTH,
Set