bug: query could't be killed

[FANNG1]

Search before asking

• I had searched in the issues and found no similar issues.

Version

5f0ea05

What's Wrong?

set max_execute_time to 10min, but a query runs serval hours, and kill query command takes no effect.

How to Reproduce?

1. run databend with 3 nodes
2. run sql

SELECT count(*) AS sqlrewriter0
FROM. db.xxx
WHERE p_date BETWEEN '20220901' AND '20220927'
  AND agent_id IN
    (SELECT agent_id
     FROM db.xxxxx
     WHERE dt = '2022-09-27'
     GROUP BY agent_id)
GROUP BY agent_id
ORDER BY sqlrewriter0
LIMIT 100

3. kill query $query_id

databend cpu is 100%, and the thread dump is

#0  0x000055f9d71f3b72 in databend_query::pipelines::processors::transforms::hash_join::result_blocks::_$LT$impl$u20$databend_query..pipelines..processors..transforms..hash_join..join_hash_table..JoinHashTable$GT$::result_blocks::h46f9186f799dbdce ()
#1  0x000055f9d725e9c8 in _$LT$databend_query..pipelines..processors..transforms..hash_join..join_hash_table..JoinHashTable$u20$as$u20$databend_query..pipelines..processors..transforms..hash_join..hash_join_state..HashJoinState$GT$::probe::hef6a8885a95eb1c4 ()
#2  0x000055f9d77d9e48 in _$LT$databend_query..pipelines..processors..transforms..transform_hash_join..TransformHashJoinProbe$u20$as$u20$common_pipeline_core..processors..processor..Processor$GT$::process::h5858b691ce9bf194 ()
#3  0x000055f9d7bec1b8 in databend_query::pipelines::executor::executor_worker_context::ExecutorWorkerContext::execute_task::h044794c1f8c8586e ()
#4  0x000055f9d7be9dd8 in databend_query::pipelines::executor::pipeline_executor::PipelineExecutor::execute_single_thread::hd2a668c4d7c0c276 ()
#5  0x000055f9d7291d6b in _$LT$core..panic..unwind_safe..AssertUnwindSafe$LT$F$GT$$u20$as$u20$core..ops..function..FnOnce$LT$$LP$$RP$$GT$$GT$::call_once::hdcbb8474153b9076 ()
#6  0x000055f9d7be98e0 in databend_query::pipelines::executor::pipeline_executor::PipelineExecutor::execute_threads::_$u7b$$u7b$closure$u7d$$u7d$::h6e78e3ee712fb3f9 ()
#7  0x000055f9d7a744c9 in std::sys_common::backtrace::__rust_begin_short_backtrace::hb97ab8724f09bbc9 ()
#8  0x000055f9d7a85d75 in core::ops::function::FnOnce::call_once$u7b$$u7b$vtable.shim$u7d$$u7d$::hf2fa816957e0c89d ()
#9  0x000055f9d8563f73 in std::sys::unix::thread::Thread::new::thread_start::hdc948a6935722a5a ()
#10 0x00007f62dd7d3dd5 in start_thread () from /lib64/libpthread.so.0

Are you willing to submit PR?

• Yes I am willing to submit a PR!

[BohuTANG]

It looks hash join doesn't handle the processor abortion, cc @zhang2014