greenplum Explicit Redistribute Motion 代码

Posted 2020-10-22 brayden

Explicit Redistribute Motion

The Explicit Redistribute iterator moves tuples to segments explicitly specified in the segment ID column of the tuples. This is differs from a Redistribute Motion iterator, where target segments are indirectly specified through hash expressions.

testdb=# explain delete from t1  where t1.id in (select id from t2);
                                                          QUERY PLAN                                                           
-------------------------------------------------------------------------------------------------------------------------------
 Delete (slice0; segments: 6)  (rows=4279005 width=10)
   ->  Explicit Redistribute Motion 6:6  (slice3; segments: 6)  (cost=57114.23..338731.20 rows=4279005 width=10)
         ->  Hash Join  (cost=57114.23..338731.20 rows=4279005 width=10)
               Hash Cond: t1.id = t2.id
               ->  Append-only Scan on t1  (cost=0.00..16880.00 rows=266667 width=14)
               ->  Hash  (cost=55868.87..55868.87 rows=16605 width=4)
                     ->  Broadcast Motion 6:6  (slice2; segments: 6)  (cost=52880.00..55868.87 rows=16605 width=4)
                           ->  HashAggregate  (cost=52880.00..53876.29 rows=16605 width=4)
                                 Group By: t2.id
                                 ->  Redistribute Motion 6:6  (slice1; segments: 6)  (cost=0.00..48880.00 rows=266667 width=4)
                                       Hash Key: t2.id
                                       ->  Append-only Scan on t2  (cost=0.00..16880.00 rows=266667 width=4)
 Optimizer status: legacy query optimizer
(13 rows)

 

1. 做plan时，调用preprocess_targetlist， 对update，delete，target list中加入 ctid和segment id；

(gdb) bt
#0  preprocess_targetlist (root=0x2304f58, tlist=0x0) at preptlist.c:152
#1  0x00000000007e868c in grouping_planner (root=0x2304f58, tuple_fraction=0) at planner.c:1814
#2  0x00000000007e79d7 in subquery_planner (glob=0x224cf88, parse=0x2304c00, parent_root=0x0, tuple_fraction=0, subroot=0x7ffffaf9a9d0, 
    config=0x2304f08) at planner.c:1196
#3  0x00000000007e6d94 in standard_planner (parse=0x2304c00, cursorOptions=0, boundParams=0x0) at planner.c:785
#4  0x00000000007e69d0 in resource_negotiator (parse=0x21feca8, cursorOptions=0, boundParams=0x0, resourceLife=QRL_ONCE, result=0x7ffffaf9ac08)
    at planner.c:658
#5  0x00000000007e62b4 in planner (parse=0x21feca8, cursorOptions=0, boundParams=0x0, resourceLife=QRL_ONCE) at planner.c:473
#6  0x000000000087c610 in pg_plan_query (querytree=0x21feca8, boundParams=0x0, resource_life=QRL_ONCE) at postgres.c:937
#7  0x000000000087c71b in pg_plan_queries (querytrees=0x224c560, boundParams=0x0, needSnapshot=0 ‘\000‘, resource_life=QRL_ONCE) at postgres.c:1011
#8  0x000000000087db9a in exec_simple_query (query_string=0x21fdc90 "delete from t1  where t1.id in (select id from t2);", seqServerHost=0x0, 
    seqServerPort=-1) at postgres.c:1771
#9  0x0000000000885033 in exec_simple_query_message (firstchar=81 ‘Q‘, input_message=0x7ffffaf9b0d0) at postgres.c:6314
#10 0x00000000008829b7 in PostgresMain (argc=4, argv=0x20c5cf8, username=0x20c5c10 "brayden") at postgres.c:4873
#11 0x000000000083615a in BackendRun (port=0x2089120) at postmaster.c:5889
#12 0x000000000083560f in BackendStartup (port=0x2089120) at postmaster.c:5484
#13 0x000000000082ff00 in ServerLoop () at postmaster.c:2163
#14 0x000000000082efcb in PostmasterMain (argc=9, argv=0x2080a40) at postmaster.c:1454
#15 0x000000000075a7ef in main (argc=9, argv=0x2080a40) at main.c:226

2. 接着plan， 加入explicit motion：

(gdb) bt
#0  request_explicit_motion (plan=0x226ad50, resultRelationsIdx=1, rtable=0x22e5be8) at cdbmutate.c:1430
#1  0x0000000000a621c7 in apply_motion (root=0x2304f58, plan=0x226ad50, query=0x2304c00) at cdbmutate.c:686
#2  0x0000000000a58c2a in cdbparallelize (root=0x2304f58, plan=0x226ad50, query=0x2304c00, cursorOptions=0, boundParams=0x0) at cdbllize.c:285
#3  0x00000000007e6f33 in standard_planner (parse=0x2304c00, cursorOptions=0, boundParams=0x0) at planner.c:834
#4  0x00000000007e69d0 in resource_negotiator (parse=0x21feca8, cursorOptions=0, boundParams=0x0, resourceLife=QRL_ONCE, result=0x7ffffaf9ac08)
    at planner.c:658
#5  0x00000000007e62b4 in planner (parse=0x21feca8, cursorOptions=0, boundParams=0x0, resourceLife=QRL_ONCE) at planner.c:473
#6  0x000000000087c610 in pg_plan_query (querytree=0x21feca8, boundParams=0x0, resource_life=QRL_ONCE) at postgres.c:937
#7  0x000000000087c71b in pg_plan_queries (querytrees=0x224c560, boundParams=0x0, needSnapshot=0 ‘\000‘, resource_life=QRL_ONCE) at postgres.c:1011
#8  0x000000000087db9a in exec_simple_query (query_string=0x21fdc90 "delete from t1  where t1.id in (select id from t2);", seqServerHost=0x0, 
    seqServerPort=-1) at postgres.c:1771
#9  0x0000000000885033 in exec_simple_query_message (firstchar=81 ‘Q‘, input_message=0x7ffffaf9b0d0) at postgres.c:6314
#10 0x00000000008829b7 in PostgresMain (argc=4, argv=0x20c5cf8, username=0x20c5c10 "brayden") at postgres.c:4873
#11 0x000000000083615a in BackendRun (port=0x2089120) at postmaster.c:5889
#12 0x000000000083560f in BackendStartup (port=0x2089120) at postmaster.c:5484
#13 0x000000000082ff00 in ServerLoop () at postmaster.c:2163
#14 0x000000000082efcb in PostmasterMain (argc=9, argv=0x2080a40) at postmaster.c:1454
#15 0x000000000075a7ef in main (argc=9, argv=0x2080a40) at main.c:226

3. 执行， 在做projection的时候， 记录 ctid， segment id：

(gdb) bt
#0  GetQEIndex () at identity.c:446
#1  0x00000000006f1404 in slot_getsysattr (slot=0x2bfc0f8, attnum=-8, isnull=0x2c4a0e9 "") at execTuples.c:1495
#2  0x00000000006e361a in slot_getattr (slot=0x2bfc0f8, attnum=-8, isnull=0x2c4a0e9 "") at ../../../src/include/executor/tuptable.h:320
#3  0x00000000006e4221 in ExecEvalVar (exprstate=0x2c49270, econtext=0x2bc7190, isNull=0x2c4a0e9 "", isDone=0x2c4967c) at execQual.c:674
#4  0x00000000006edd84 in ExecTargetList (targetlist=0x2c49210, econtext=0x2bc7190, values=0x2c4a0b8, isnull=0x2c4a0e8 "", itemIsDone=0x2c49678, 
    isDone=0x0) at execQual.c:5565
#5  0x00000000006ee300 in ExecProject (projInfo=0x2c495e8, isDone=0x0) at execQual.c:5762
#6  0x00000000006ee808 in ExecScan (node=0x2c48d10, accessMtd=0x6f65b4 <AppendOnlyScanNext>) at execScan.c:187
#7  0x00000000006eed3c in ExecTableScanRelation (scanState=0x2c48d10) at execScan.c:460
#8  0x000000000071e703 in ExecTableScan (node=0x2c48d10) at nodeTableScan.c:59
#9  0x00000000006e1d5e in ExecProcNode (node=0x2c48d10) at execProcnode.c:1158
#10 0x000000000070a15a in ExecHashJoinOuterGetTuple (outerNode=0x2c48d10, hjstate=0x2bc2a40, hashvalue=0x7ffda17d2264) at nodeHashjoin.c:822
#11 0x000000000070956a in ExecHashJoin (node=0x2bc2a40) at nodeHashjoin.c:362
#12 0x00000000006e1e89 in ExecProcNode (node=0x2bc2a40) at execProcnode.c:1217
#13 0x0000000000711c45 in execMotionSender (node=0x2bcc5b8) at nodeMotion.c:364
#14 0x0000000000711c03 in ExecMotion (node=0x2bcc5b8) at nodeMotion.c:331
#15 0x00000000006e1f31 in ExecProcNode (node=0x2bcc5b8) at execProcnode.c:1249
#16 0x00000000006dbba2 in ExecutePlan (estate=0x2bcaa40, planstate=0x2bcc5b8, operation=CMD_SELECT, numberTuples=0, direction=ForwardScanDirection, 
    dest=0x2c35c90) at execMain.c:3298
#17 0x00000000006d855f in ExecutorRun (queryDesc=0x2b9b210, direction=ForwardScanDirection, count=0) at execMain.c:1217
#18 0x000000000088713d in ProcessQuery (portal=0x2b990d0, stmt=0x2c21350, params=0x0, dest=0x2c35c90, completionTag=0x7ffda17d2a90 "")
    at pquery.c:327
#19 0x000000000088a0db in PortalRunMulti (portal=0x2b990d0, isTopLevel=1 ‘\001‘, dest=0x2c35c90, altdest=0x2c35c90, completionTag=0x7ffda17d2a90 "")
    at pquery.c:2061
#20 0x0000000000889578 in PortalRun (portal=0x2b990d0, count=9223372036854775807, isTopLevel=1 ‘\001‘, dest=0x2c35c90, altdest=0x2c35c90, 
    completionTag=0x7ffda17d2a90 "") at pquery.c:1599
#21 0x000000000087d5a4 in exec_mpp_query (query_string=0x2b750b2 "delete from t1  where t1.id in (select id from t2);", serializedQuerytree=0x0, 
    serializedQuerytreelen=0, serializedPlantree=0x2b750e6 "+2", serializedPlantreelen=2187, serializedParams=0x0, serializedParamslen=0, 
    serializedSliceInfo=0x2b75971 "\267\003", serializedSliceInfolen=302, serializedResource=0x2b75aec "P", serializedResourceLen=41, 
    seqServerHost=0x2b75b15 "192.168.3.175", seqServerPort=18040, localSlice=3) at postgres.c:1508
#22 0x0000000000883128 in PostgresMain (argc=266, argv=0x2a413b0, username=0x2a22c10 "brayden") at postgres.c:5057
#23 0x000000000083615a in BackendRun (port=0x29e2520) at postmaster.c:5889
#24 0x000000000083560f in BackendStartup (port=0x29e2520) at postmaster.c:5484
#25 0x000000000082ff00 in ServerLoop () at postmaster.c:2163
---Type <return> to continue, or q <return> to quit--- 
#26 0x000000000082efcb in PostmasterMain (argc=9, argv=0x29dda40) at postmaster.c:1454
#27 0x000000000075a7ef in main (argc=9, argv=0x29dda40) at main.c:226
(gdb) 

4. 执行， motion send tuple， 发送给对应的sement：

(gdb) bt
#0  doSendTuple (motion=0x2c21560, node=0x2bcc5b8, outerTupleSlot=0x2bfc178) at nodeMotion.c:1620
#1  0x0000000000711cbc in execMotionSender (node=0x2bcc5b8) at nodeMotion.c:400
#2  0x0000000000711c03 in ExecMotion (node=0x2bcc5b8) at nodeMotion.c:331
#3  0x00000000006e1f31 in ExecProcNode (node=0x2bcc5b8) at execProcnode.c:1249
#4  0x00000000006dbba2 in ExecutePlan (estate=0x2bcaa40, planstate=0x2bcc5b8, operation=CMD_SELECT, numberTuples=0, direction=ForwardScanDirection, 
    dest=0x2c35c90) at execMain.c:3298
#5  0x00000000006d855f in ExecutorRun (queryDesc=0x2b9b210, direction=ForwardScanDirection, count=0) at execMain.c:1217
#6  0x000000000088713d in ProcessQuery (portal=0x2b990d0, stmt=0x2c21350, params=0x0, dest=0x2c35c90, completionTag=0x7ffda17d2a90 "")
    at pquery.c:327
#7  0x000000000088a0db in PortalRunMulti (portal=0x2b990d0, isTopLevel=1 ‘\001‘, dest=0x2c35c90, altdest=0x2c35c90, completionTag=0x7ffda17d2a90 "")
    at pquery.c:2061
#8  0x0000000000889578 in PortalRun (portal=0x2b990d0, count=9223372036854775807, isTopLevel=1 ‘\001‘, dest=0x2c35c90, altdest=0x2c35c90, 
    completionTag=0x7ffda17d2a90 "") at pquery.c:1599
#9  0x000000000087d5a4 in exec_mpp_query (query_string=0x2b750b2 "delete from t1  where t1.id in (select id from t2);", serializedQuerytree=0x0, 
    serializedQuerytreelen=0, serializedPlantree=0x2b750e6 "+2", serializedPlantreelen=2187, serializedParams=0x0, serializedParamslen=0, 
    serializedSliceInfo=0x2b75971 "\267\003", serializedSliceInfolen=302, serializedResource=0x2b75aec "P", serializedResourceLen=41, 
    seqServerHost=0x2b75b15 "192.168.3.175", seqServerPort=18040, localSlice=3) at postgres.c:1508
#10 0x0000000000883128 in PostgresMain (argc=266, argv=0x2a413b0, username=0x2a22c10 "brayden") at postgres.c:5057
#11 0x000000000083615a in BackendRun (port=0x29e2520) at postmaster.c:5889
#12 0x000000000083560f in BackendStartup (port=0x29e2520) at postmaster.c:5484
#13 0x000000000082ff00 in ServerLoop () at postmaster.c:2163
#14 0x000000000082efcb in PostmasterMain (argc=9, argv=0x29dda40) at postmaster.c:1454
#15 0x000000000075a7ef in main (argc=9, argv=0x29dda40) at main.c:226