doris中rollup表没有分区？

[liumengkai]

我用不带分区过滤的sql explain的时候显示用的新的roll up
但是where 中加上时间day的过滤 explain 就显示还是用的base rollup index 就是旧的

下面是两组sql运行explain结果

explain select code,sum(pv) from test_sum where day > '2022-05-01' group by code
explain select code,sum(pv) from test_sum group by code

Result:

day 是我的分区字段
test_sum 是我的base rollup index 也就是默认的
rollup_userid是我创建的rollup index . rollup_userid(code,pv)

explain select code,sum(pv) from test_sum where day > '2022-05-01' group by code:
+---------------------------------------------+
| Explain String |
+---------------------------------------------+
| PLAN FRAGMENT 0 |
| OUTPUT EXPRS:3 | 5: sum |
| PARTITION: UNPARTITIONED |
| |
| RESULT SINK |
| |
| 6:EXCHANGE |
| |
| PLAN FRAGMENT 1 |
| OUTPUT EXPRS: |
| PARTITION: HASH_PARTITIONED: 6: code |
| |
| STREAM DATA SINK |
| EXCHANGE ID: 06 |
| UNPARTITIONED |
| |
| 5:Decode |
| | <dict id 6> : <string id 3> |
| | |
| 4:AGGREGATE (merge finalize) |
| | output: sum(5: sum) |
| | group by: 6: code |
| | |
| 3:EXCHANGE |
| |
| PLAN FRAGMENT 2 |
| OUTPUT EXPRS: |
| PARTITION: RANDOM |
| |
| STREAM DATA SINK |
| EXCHANGE ID: 03 |
| HASH_PARTITIONED: 6: code |
| |
| 2:AGGREGATE (update serialize) |
| | STREAMING |
| | output: sum(4: pv) |
| | group by: 6: code |
| | |
| 1:Project |
| | <slot 4> : 4: pv |
| | <slot 6> : 6: code |
| | |
| 0:OlapScanNode |
| TABLE: test_sum |
| PREAGGREGATION: ON |
| PREDICATES: 2: day > '2022-05-01' |
| partitions=1/3 |
| rollup: test_sum |
| tabletRatio=4/4 |
| tabletList=455005,455009,455013,455017 |
| cardinality=0 |
| avgRowSize=13.0 |
| numNodes=0 |
+---------------------------------------------+

rollup: test_sum

explain select code,sum(pv) from test_sum group by code:

+-------------------------------------------------------------------------+
| Explain String |
+-------------------------------------------------------------------------+
| PLAN FRAGMENT 0 |
| OUTPUT EXPRS:3 | 5: sum |
| PARTITION: UNPARTITIONED |
| |
| RESULT SINK |
| |
| 5:EXCHANGE |
| |
| PLAN FRAGMENT 1 |
| OUTPUT EXPRS: |
| PARTITION: HASH_PARTITIONED: 6: code |
| |
| STREAM DATA SINK |
| EXCHANGE ID: 05 |
| UNPARTITIONED |
| |
| 4:Decode |
| | <dict id 6> : <string id 3> |
| | |
| 3:AGGREGATE (merge finalize) |
| | output: sum(5: sum) |
| | group by: 6: code |
| | |
| 2:EXCHANGE |
| |
| PLAN FRAGMENT 2 |
| OUTPUT EXPRS: |
| PARTITION: RANDOM |
| |
| STREAM DATA SINK |
| EXCHANGE ID: 02 |
| HASH_PARTITIONED: 6: code |
| |
| 1:AGGREGATE (update serialize) |
| | STREAMING |
| | output: sum(4: pv) |
| | group by: 6: code |
| | |
| 0:OlapScanNode |
| TABLE: test_sum |
| PREAGGREGATION: ON |
| partitions=2/3 |
| rollup: rollup_userid |
| tabletRatio=8/8 |
| tabletList=455279,455283,455287,455291,455263,455267,455271,455275 |
| cardinality=2 |
| avgRowSize=9.0 |
| numNodes=0 |
+-------------------------------------------------------------------------+

rollup: rollup_userid

这样是正常的吗？也就是说用上分区字段就不能中新的roll up吗？

[liumengkai]

因为创建rollup的时候没有把分区字段加进去:-)