mirror of
https://github.com/timescale/timescaledb.git
synced 2025-05-19 04:03:06 +08:00
2b8f98c616
If a lot of chunks are involved then the current pl/pgsql function to compute the size of each chunk via a nested loop is pretty slow. Additionally, the current functionality makes a system call to get the file size on disk for each chunk everytime this function is called. That again slows things down. We now have an approximate function which is implemented in C to avoid the issues in the pl/pgsql function. Additionally, this function also uses per backend caching using the smgr layer to compute the approximate size cheaply. The PG cache invalidation clears off the cached size for a chunk when DML happens into it. That size cache is thus able to get the latest size in a matter of minutes. Also, due to the backend caching, any long running session will only fetch latest data for new or modified chunks and can use the cached data (which is calculated afresh the first time around) effectively for older chunks.
74 lines
3.7 KiB
SQL
74 lines
3.7 KiB
SQL
-- This file and its contents are licensed under the Timescale License.
|
|
-- Please see the included NOTICE for copyright information and
|
|
-- LICENSE-TIMESCALE for a copy of the license.
|
|
|
|
--------------------------------------------------------------------------
|
|
-- show_chunks and drop_chunks functions on a compressed table
|
|
-- (issue https://github.com/timescale/timescaledb/issues/1535)
|
|
|
|
-- create a table that will not be compressed
|
|
CREATE TABLE public.uncompressed_table(time date NOT NULL, temp float8, device_id text);
|
|
CREATE INDEX ON public.uncompressed_table(time DESC);
|
|
SELECT create_hypertable('public.uncompressed_table', 'time', chunk_time_interval => interval '1 day');
|
|
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-01', 1.0, 'dev1');
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-05', 2.0, 'dev1');
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-07', 3.0, 'dev1');
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-08', 4.0, 'dev7');
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-09', 5.0, 'dev7');
|
|
INSERT INTO public.uncompressed_table VALUES('2020-03-10', 6.0, 'dev7');
|
|
|
|
-- create next table that is going to be compressed:
|
|
CREATE TABLE public.table_to_compress (time date NOT NULL, acq_id bigint, value bigint);
|
|
CREATE INDEX idx_table_to_compress_acq_id ON public.table_to_compress(acq_id);
|
|
SELECT create_hypertable('public.table_to_compress', 'time', chunk_time_interval => interval '1 day');
|
|
ALTER TABLE public.table_to_compress SET (timescaledb.compress, timescaledb.compress_segmentby = 'acq_id');
|
|
INSERT INTO public.table_to_compress VALUES ('2020-01-01', 1234567, 777888);
|
|
INSERT INTO public.table_to_compress VALUES ('2020-02-01', 567567, 890890);
|
|
INSERT INTO public.table_to_compress VALUES ('2020-02-10', 1234, 5678);
|
|
SELECT show_chunks('public.uncompressed_table');
|
|
SELECT show_chunks('public.table_to_compress');
|
|
SELECT show_chunks('public.table_to_compress', older_than=>'1 day'::interval);
|
|
SELECT show_chunks('public.table_to_compress', newer_than=>'1 day'::interval);
|
|
-- compress all chunks of the table:
|
|
SELECT compress_chunk(show_chunks('public.table_to_compress'));
|
|
-- check that approx size function works. We call VACUUM to ensure all forks exist
|
|
VACUUM public.table_to_compress;
|
|
SELECT * FROM hypertable_approximate_size('public.table_to_compress');
|
|
SELECT * FROM hypertable_size('public.table_to_compress');
|
|
-- and run the queries again to make sure results are the same
|
|
SELECT show_chunks('public.uncompressed_table');
|
|
SELECT show_chunks('public.table_to_compress');
|
|
SELECT show_chunks('public.table_to_compress', older_than=>'1 day'::interval);
|
|
SELECT show_chunks('public.table_to_compress', newer_than=>'1 day'::interval);
|
|
-- truncate one compressed chunk
|
|
SELECT chunk_schema || '.' || chunk_name as "CHNAME"
|
|
FROM timescaledb_information.chunks
|
|
WHERE hypertable_name = 'table_to_compress' and hypertable_schema = 'public'
|
|
ORDER BY chunk_name LIMIT 1
|
|
\gset
|
|
SELECT count(*) FROM :CHNAME;
|
|
TRUNCATE TABLE :CHNAME;
|
|
SELECT count(*) FROM :CHNAME;
|
|
-- drop all hypertables' old chunks
|
|
SELECT drop_chunks(table_name::regclass, older_than=>'1 day'::interval)
|
|
FROM _timescaledb_catalog.hypertable
|
|
WHERE schema_name = current_schema()
|
|
ORDER BY table_name DESC;
|
|
SELECT show_chunks('public.uncompressed_table');
|
|
SELECT show_chunks('public.table_to_compress');
|
|
|
|
-- test calling on internal compressed table
|
|
SELECT
|
|
format('%I.%I', ht.schema_name, ht.table_name) AS "TABLENAME"
|
|
FROM
|
|
_timescaledb_catalog.hypertable ht
|
|
INNER JOIN _timescaledb_catalog.hypertable uncompress ON (ht.id = uncompress.compressed_hypertable_id
|
|
AND uncompress.table_name = 'table_to_compress') \gset
|
|
|
|
\set ON_ERROR_STOP 0
|
|
SELECT show_chunks(:'TABLENAME');
|
|
SELECT drop_chunks(:'TABLENAME',now());
|
|
\set ON_ERROR_STOP 1
|
|
|