Flink SQL, How can i set parallelism in clause of group by ?

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Flink SQL, How can i set parallelism in clause of group by ?

forideal
Hello everyone

     I'm a Flink SQL user. Now i have a question.How can i set parallelism in clause of group by.
     For example
     SELECT 
            T.user_id,
            D.user_name
     (SELECT 
             user_id,
             MIN(processtime)
     from my_table 
    group by user_id,TUMBLE(processtime, INTERVAL '15' SECOND)) AS T
     LEFT JOIN my_dim_table FOR SYSTEM_TIME AS OF T.proc_time AS D ON T.user_id = D.user_id
     
     If the partition of the topic is 3, then,the parallelism of the job is 3.
     In this example,there are three operator
         0ne is Source operator, parallelism is 3
         Two is GroupWindowAggregate operator,parallelism is 3
         Three is LookupJoin operator,parallelism is 3

      I want to change the parallelism of GroupWindowAggregate,but i can't.

Best wishes
forideal


 

Reply | Threaded
Open this post in threaded view
|

Re: Flink SQL, How can i set parallelism in clause of group by ?

Benchao Li
Hi forideal,

Currently, there is no way to change an operator's parallelism for SQL. All operators use global parallelism as their parallelism, except for some 'global operator' which can only has parallelism of 1.

BTW, does changing the parallelism of all operators meets your need?

forideal <[hidden email]> 于2020年3月18日周三 上午12:50写道:
Hello everyone

     I'm a Flink SQL user. Now i have a question.How can i set parallelism in clause of group by.
     For example
     SELECT 
            T.user_id,
            D.user_name
     (SELECT 
             user_id,
             MIN(processtime)
     from my_table 
    group by user_id,TUMBLE(processtime, INTERVAL '15' SECOND)) AS T
     LEFT JOIN my_dim_table FOR SYSTEM_TIME AS OF T.proc_time AS D ON T.user_id = D.user_id
     
     If the partition of the topic is 3, then,the parallelism of the job is 3.
     In this example,there are three operator
         0ne is Source operator, parallelism is 3
         Two is GroupWindowAggregate operator,parallelism is 3
         Three is LookupJoin operator,parallelism is 3

      I want to change the parallelism of GroupWindowAggregate,but i can't.

Best wishes
forideal


 



--
Benchao Li
School of Electronics Engineering and Computer Science, Peking University
Tel:+86-15650713730
Email: [hidden email]; [hidden email]