[Enhancement]: Concurrent drop_chunk()

[mkindahl]

What type of enhancement is this?

Performance

What subsystems and features will be improved?

Command processing

What does the enhancement do?

When dropping a chunk from a hypertable using drop_chunks(), it will take quite heavy locks to be able to drop the chunk:

mats=# start transaction;
START TRANSACTION
mats=*# select drop_chunks('readings', older_than => '2022-06-14');
              drop_chunks               
----------------------------------------
 _timescaledb_internal._hyper_2_1_chunk
 _timescaledb_internal._hyper_2_2_chunk
(2 rows)

Locks taken are:

mats=# select relation::regclass, mode from pg_locks join pg_stat_activity using (pid) where locktype = 'relation' and pid = 1074774;
                      relation                       |        mode         
-----------------------------------------------------+---------------------
 _timescaledb_catalog.dimension_slice                | RowShareLock
 _timescaledb_catalog.dimension_slice                | RowExclusiveLock
 _timescaledb_internal._hyper_2_2_chunk              | AccessExclusiveLock
 _timescaledb_internal."1_1_readings_created_at_key" | AccessExclusiveLock
 _timescaledb_internal."2_2_readings_created_at_key" | AccessExclusiveLock
 _timescaledb_internal._hyper_2_1_chunk              | AccessExclusiveLock
(6 rows)

Having a concurrent version of drop_chunks() that take locks similar to locking done by ALTER TABLE ... DETACH PARTITION (that is SHARE UPDATE EXCLUSIVE) would avoid blocking readers and writers of the hypertable.

Implementation challenges

No response