限制查询大小以控制事务日志大小

Question

This query causes our transaction log to grow to 25GB. The database is in SIMPLE mode.

INSERT  INTO updbl.dbo.PopulationRelatives
    ( personid,
      personsex,
      relativeid,
      relativesex,
      degree,
      relationship,
      maternalpaternal )
    SELECT DISTINCT
            personid = relative1,
            relative1sex,
            relative2,
            relative2sex,
            degree,
            relationship = Rel1Rel2,
            maternalpaternal
    FROM    UPDBwork.dbo.DegreeRelationship

By looping I was able to limit the growth to 8GB.

SELECT @PID = 0, @BatchSize = 1000000, @ROWCOUNT = 0
SELECT @MaxPID = MAX(relative1) FROM updbwork.dbo.DegreeRelationship
WHILE @PID < @MaxPID+@BatchSize
BEGIN
INSERT  INTO updbl.dbo.PopulationRelatives
        ( personid,
          personsex,
          relativeid,
          relativesex,
          degree,
          relationship,
          maternalpaternal )
        SELECT DISTINCT
                personid = relative1,
                relative1sex,
                relative2,
                relative2sex,
                degree,
                relationship = Rel1Rel2,
                maternalpaternal
        FROM    UPDBwork.dbo.DegreeRelationship
        WHERE relative1 BETWEEN @PID+1 AND @PID+@BatchSize
        SET @PID = @PID + @BatchSize
CHECKPOINT
END

This isn't the best strategy as each loop produces a different number of rows depending on the DISTINCT values. Unfortunately there is no good ID to partition the data on. Is there some way I could control for the size of each group? I was thinking of adding TOP(X) but the engine would still have to do a large calculation to satisfy the DISTINCT statement. A cursor would be great but again, how do I find my DISTINCT values? I am just hoping for some brain storming here. Thanks.

Answer 1

Sounds like bulk operation... if changing the recovery model is an option temporarily change it to bulk-logged. Here is a link that may be of help: http://technet.microsoft.com/en-us/library/ms175987(v=SQL.105).aspx