When I run select 1 query in 20 node cluster. It takes 30 seconds and out of which 29 seconds are spent on Before plan phase. Does that mean we do have a network issue? How to debug this.
It may or may not be a network issue. Please profile the query and review the output of execution_engine_profiles to check which operator is taking long time.
How did you check select 1 took 29 seconds on Before plan phase?
You can find where the time is spend with this query :
select time, node_name, execution_step, completion_time, completion_time - time as Step_Duration from dc_query_executions where transaction_id=<> and statement_id=<< statement_id>> order by 2,1 ;
Can you confirm that is in the prepare_plan phase? and in the initiator?