timescaledb/sql/updates/latest-dev.sql

-- Remove multi-node CAGG support
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_cagg_log_add_entry(integer,bigint,bigint);
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_hyper_log_add_entry(integer,bigint,bigint);
DROP FUNCTION IF EXISTS _timescaledb_internal.materialization_invalidation_log_delete(integer);
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_process_cagg_log(integer,integer,regtype,bigint,bigint,integer[],bigint[],bigint[]);
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_process_cagg_log(integer,integer,regtype,bigint,bigint,integer[],bigint[],bigint[],text[]);
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_process_hypertable_log(integer,integer,regtype,integer[],bigint[],bigint[]);
DROP FUNCTION IF EXISTS _timescaledb_internal.invalidation_process_hypertable_log(integer,integer,regtype,integer[],bigint[],bigint[],text[]);
DROP FUNCTION IF EXISTS _timescaledb_internal.hypertable_invalidation_log_delete(integer);

DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_cagg_log_add_entry(integer,bigint,bigint);
DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_hyper_log_add_entry(integer,bigint,bigint);
DROP FUNCTION IF EXISTS _timescaledb_functions.materialization_invalidation_log_delete(integer);
DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_process_cagg_log(integer,integer,regtype,bigint,bigint,integer[],bigint[],bigint[]);
DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_process_cagg_log(integer,integer,regtype,bigint,bigint,integer[],bigint[],bigint[],text[]);
DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_process_hypertable_log(integer,integer,regtype,integer[],bigint[],bigint[]);
DROP FUNCTION IF EXISTS _timescaledb_functions.invalidation_process_hypertable_log(integer,integer,regtype,integer[],bigint[],bigint[],text[]);
DROP FUNCTION IF EXISTS _timescaledb_functions.hypertable_invalidation_log_delete(integer);

-- Remove chunk metadata when marked as dropped
CREATE FUNCTION _timescaledb_functions.remove_dropped_chunk_metadata(_hypertable_id INTEGER)
RETURNS INTEGER LANGUAGE plpgsql AS $$
DECLARE
  _chunk_id INTEGER;
  _removed INTEGER := 0;
BEGIN
  FOR _chunk_id IN
    SELECT id FROM _timescaledb_catalog.chunk
    WHERE hypertable_id = _hypertable_id
    AND dropped IS TRUE
    AND NOT EXISTS (
        SELECT FROM information_schema.tables
        WHERE tables.table_schema = chunk.schema_name
        AND tables.table_name = chunk.table_name
    )
    AND NOT EXISTS (
        SELECT FROM _timescaledb_catalog.hypertable
        JOIN _timescaledb_catalog.continuous_agg ON continuous_agg.raw_hypertable_id = hypertable.id
        WHERE hypertable.id = chunk.hypertable_id
        -- for the old caggs format we need to keep chunk metadata for dropped chunks
        AND continuous_agg.finalized IS FALSE
    )
  LOOP
    _removed := _removed + 1;
    RAISE INFO 'Removing metadata of chunk % from hypertable %', _chunk_id, _hypertable_id;
    WITH _dimension_slice_remove AS (
        DELETE FROM _timescaledb_catalog.dimension_slice
        USING _timescaledb_catalog.chunk_constraint
        WHERE dimension_slice.id = chunk_constraint.dimension_slice_id
        AND chunk_constraint.chunk_id = _chunk_id
        RETURNING _timescaledb_catalog.dimension_slice.id
    )
    DELETE FROM _timescaledb_catalog.chunk_constraint
    USING _dimension_slice_remove
    WHERE chunk_constraint.dimension_slice_id = _dimension_slice_remove.id;

    DELETE FROM _timescaledb_internal.bgw_policy_chunk_stats
    WHERE bgw_policy_chunk_stats.chunk_id = _chunk_id;

    DELETE FROM _timescaledb_catalog.chunk_index
    WHERE chunk_index.chunk_id = _chunk_id;

    DELETE FROM _timescaledb_catalog.compression_chunk_size
    WHERE compression_chunk_size.chunk_id = _chunk_id
    OR compression_chunk_size.compressed_chunk_id = _chunk_id;

    DELETE FROM _timescaledb_catalog.chunk
    WHERE chunk.id = _chunk_id
    OR chunk.compressed_chunk_id = _chunk_id;
  END LOOP;

  RETURN _removed;
END;
$$ SET search_path TO pg_catalog, pg_temp;

SELECT _timescaledb_functions.remove_dropped_chunk_metadata(id) AS chunks_metadata_removed
FROM _timescaledb_catalog.hypertable;

--
-- Rebuild the catalog table `_timescaledb_catalog.continuous_aggs_bucket_function`
--

CREATE OR REPLACE FUNCTION _timescaledb_functions.cagg_get_bucket_function(
    mat_hypertable_id INTEGER
) RETURNS regprocedure AS '@MODULE_PATHNAME@', 'ts_continuous_agg_get_bucket_function' LANGUAGE C STRICT VOLATILE;

-- Since we need now the regclass of the used bucket function, we have to recover it
-- by parsing the view query by calling 'cagg_get_bucket_function'.
CREATE TABLE _timescaledb_catalog._tmp_continuous_aggs_bucket_function AS
    SELECT
      mat_hypertable_id,
      _timescaledb_functions.cagg_get_bucket_function(mat_hypertable_id),
      bucket_width,
      origin,
      NULL::text AS bucket_offset,
      timezone,
      false AS bucket_fixed_width
    FROM
      _timescaledb_catalog.continuous_aggs_bucket_function
    ORDER BY
         mat_hypertable_id;

ALTER EXTENSION timescaledb
    DROP TABLE _timescaledb_catalog.continuous_aggs_bucket_function;

DROP TABLE _timescaledb_catalog.continuous_aggs_bucket_function;

CREATE TABLE _timescaledb_catalog.continuous_aggs_bucket_function (
  mat_hypertable_id integer NOT NULL,
  -- The bucket function
  bucket_func regprocedure NOT NULL,
  -- `bucket_width` argument of the function, e.g. "1 month"
  bucket_width text NOT NULL,
  -- optional `origin` argument of the function provided by the user
  bucket_origin text,
  -- optional `offset` argument of the function provided by the user
  bucket_offset text,
  -- optional `timezone` argument of the function provided by the user
  bucket_timezone text,
  -- fixed or variable sized bucket
  bucket_fixed_width bool NOT NULL,
  -- table constraints
  CONSTRAINT continuous_aggs_bucket_function_pkey PRIMARY KEY (mat_hypertable_id),
  CONSTRAINT continuous_aggs_bucket_function_mat_hypertable_id_fkey FOREIGN KEY (mat_hypertable_id) REFERENCES _timescaledb_catalog.hypertable (id) ON DELETE CASCADE
);

INSERT INTO _timescaledb_catalog.continuous_aggs_bucket_function
  SELECT * FROM _timescaledb_catalog._tmp_continuous_aggs_bucket_function;

DROP TABLE _timescaledb_catalog._tmp_continuous_aggs_bucket_function;

SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_aggs_bucket_function', '');

GRANT SELECT ON TABLE _timescaledb_catalog.continuous_aggs_bucket_function TO PUBLIC;

ANALYZE _timescaledb_catalog.continuous_aggs_bucket_function;

ALTER EXTENSION timescaledb DROP FUNCTION _timescaledb_functions.cagg_get_bucket_function(INTEGER);
DROP FUNCTION IF EXISTS _timescaledb_functions.cagg_get_bucket_function(INTEGER);

--
-- End rebuild the catalog table `_timescaledb_catalog.continuous_aggs_bucket_function`
--

-- Convert _timescaledb_catalog.continuous_aggs_bucket_function.bucket_origin to TimestampTZ
UPDATE _timescaledb_catalog.continuous_aggs_bucket_function
   SET bucket_origin = bucket_origin::timestamp::timestamptz::text
   WHERE length(bucket_origin) > 1;

-- Historically, we have used empty strings for undefined bucket_origin and timezone
-- attributes. This is now replaced by proper NULL values. We use TRIM() to ensure we handle empty string well.
UPDATE _timescaledb_catalog.continuous_aggs_bucket_function SET bucket_origin = NULL WHERE TRIM(bucket_origin) = '';
UPDATE _timescaledb_catalog.continuous_aggs_bucket_function SET bucket_timezone = NULL WHERE TRIM(bucket_timezone) = '';

-- So far, there were no difference between 0 and -1 retries. Since now on, 0 means no retries. Updating the retry
-- count of existing jobs to -1 to keep the current semantics.
UPDATE _timescaledb_config.bgw_job SET max_retries = -1 WHERE max_retries = 0;

DROP FUNCTION IF EXISTS _timescaledb_functions.get_chunk_relstats;
DROP FUNCTION IF EXISTS _timescaledb_functions.get_chunk_colstats;
DROP FUNCTION IF EXISTS _timescaledb_internal.get_chunk_relstats;
DROP FUNCTION IF EXISTS _timescaledb_internal.get_chunk_colstats;

-- In older TSDB versions, we disabled autovacuum for compressed chunks
-- to keep the statistics. However, this restriction was removed in
-- #5118 but no migration was performed to remove the custom
-- autovacuum setting for existing chunks.
DO $$
DECLARE
  chunk regclass;
BEGIN
  FOR chunk IN
  SELECT pg_catalog.format('%I.%I', schema_name, table_name)::regclass
    FROM _timescaledb_catalog.chunk WHERE dropped = false
  LOOP
    EXECUTE pg_catalog.format('ALTER TABLE %s RESET (autovacuum_enabled);', chunk::text);
  END LOOP;
END
$$;