timescaledb/sql/size_utils.sql
gayyappan 7d3b4b5442 New size utils functions
Add hypertable_detailed_size , chunk_detailed_size,
hypertable_size functions.
Remove hypertable_relation_size,
hypertable_relation_size_pretty, and indexes_relation_size_pretty
Remove size information from hypertables view.
2020-07-29 15:30:39 -04:00

664 lines
23 KiB
PL/PgSQL

-- This file and its contents are licensed under the Apache License 2.0.
-- Please see the included NOTICE for copyright information and
-- LICENSE-APACHE for a copy of the license.
-- This file contains utility functions to get the relation size
-- of hypertables, chunks, and indexes on hypertables.
CREATE OR REPLACE VIEW _timescaledb_internal.hypertable_chunk_local_size AS
SELECT
h.schema_name,
h.table_name,
h.id as hypertable_id,
c.id as chunk_id,
c.schema_name as chunk_schema,
c.table_name as chunk_name,
pg_total_relation_size(format('%I.%I', c.schema_name, c.table_name))::bigint AS total_bytes,
pg_indexes_size(format('%I.%I', c.schema_name, c.table_name))::bigint AS index_bytes,
pg_total_relation_size(reltoastrelid)::bigint AS toast_bytes,
map.compressed_heap_size,
map.compressed_index_size,
map.compressed_toast_size
FROM
_timescaledb_catalog.hypertable h
INNER JOIN
_timescaledb_catalog.chunk c
ON h.id = c.hypertable_id
and c.dropped = false
INNER JOIN
pg_class pgc
ON pgc.relname = h.table_name
INNER JOIN
pg_namespace pns
ON pns.oid = pgc.relnamespace
AND pns.nspname = h.schema_name
LEFT OUTER JOIN
_timescaledb_catalog.compression_chunk_size map
ON map.chunk_id = c.id
WHERE pgc.relkind = 'r';
GRANT SELECT ON _timescaledb_internal.hypertable_chunk_local_size TO PUBLIC;
CREATE OR REPLACE FUNCTION _timescaledb_internal.data_node_hypertable_info(
node_name NAME,
schema_name_in name,
table_name_in name
)
RETURNS TABLE (
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint)
AS '@MODULE_PATHNAME@', 'ts_dist_remote_hypertable_info' LANGUAGE C VOLATILE STRICT;
CREATE OR REPLACE FUNCTION _timescaledb_internal.data_node_chunk_info(
node_name NAME,
schema_name_in name,
table_name_in name
)
RETURNS TABLE (
chunk_id integer,
chunk_schema name,
chunk_name name,
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint)
AS '@MODULE_PATHNAME@', 'ts_dist_remote_chunk_info' LANGUAGE C VOLATILE STRICT;
CREATE OR REPLACE FUNCTION _timescaledb_internal.hypertable_local_size(
schema_name_in name,
table_name_in name)
RETURNS TABLE (
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint)
LANGUAGE PLPGSQL STABLE STRICT AS
$BODY$
BEGIN
RETURN QUERY
SELECT
(sub2.table_bytes + sub2.compressed_heap_bytes)::bigint as heap_bytes,
(sub2.index_bytes + sub2.compressed_index_bytes)::bigint as index_bytes,
(sub2.toast_bytes + sub2.compressed_toast_bytes)::bigint as toast_bytes,
(sub2.total_bytes + sub2.compressed_heap_bytes + sub2.compressed_index_bytes + sub2.compressed_toast_bytes)::bigint as total_bytes
FROM
(
SELECT
*,
sub1.total_bytes - sub1.index_bytes - sub1.toast_bytes AS table_bytes
FROM
(
SELECT
sum(ch.total_bytes) as total_bytes,
COALESCE( sum(ch.index_bytes) , 0 ) as index_bytes,
COALESCE( sum(ch.toast_bytes), 0 ) as toast_bytes,
COALESCE( sum(ch.compressed_heap_size) , 0 ) as compressed_heap_bytes,
COALESCE( sum(ch.compressed_index_size) , 0) as compressed_index_bytes,
COALESCE( sum(ch.compressed_toast_size) , 0 ) as compressed_toast_bytes
FROM
_timescaledb_internal.hypertable_chunk_local_size ch
WHERE
schema_name = schema_name_in
AND table_name = table_name_in
GROUP BY
hypertable_id
) sub1
) sub2;
END;
$BODY$;
CREATE OR REPLACE FUNCTION _timescaledb_internal.hypertable_remote_size(
schema_name_in name,
table_name_in name)
RETURNS TABLE (
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint,
node_name NAME)
LANGUAGE PLPGSQL STABLE STRICT AS
$BODY$
BEGIN
RETURN QUERY
SELECT
sum(entry.table_bytes)::bigint AS table_bytes,
sum(entry.index_bytes)::bigint AS index_bytes,
sum(entry.toast_bytes)::bigint AS toast_bytes,
sum(entry.total_bytes)::bigint AS total_bytes,
srv.node_name
FROM (
SELECT
s.node_name,
_timescaledb_internal.ping_data_node (s.node_name) AS node_up
FROM
_timescaledb_catalog.hypertable AS ht,
_timescaledb_catalog.hypertable_data_node AS s
WHERE
ht.schema_name = schema_name_in
AND ht.table_name = table_name_in
AND s.hypertable_id = ht.id
) AS srv
LEFT OUTER JOIN LATERAL _timescaledb_internal.data_node_hypertable_info(
CASE WHEN srv.node_up THEN
srv.node_name
ELSE
NULL
END, schema_name_in, table_name_in) entry ON TRUE
GROUP BY srv.node_name;
END;
$BODY$;
-- Get relation size of hypertable
-- like pg_relation_size(hypertable)
--
-- main_table - hypertable to get size of
--
-- Returns:
-- table_bytes - Disk space used by main_table (like pg_relation_size(main_table))
-- index_bytes - Disk space used by indexes
-- toast_bytes - Disk space of toast tables
-- total_bytes - Total disk space used by the specified table, including all indexes and TOAST data
CREATE OR REPLACE FUNCTION hypertable_detailed_size(
main_table REGCLASS
)
RETURNS TABLE (table_bytes BIGINT,
index_bytes BIGINT,
toast_bytes BIGINT,
total_bytes BIGINT,
node_name NAME
) LANGUAGE PLPGSQL STABLE STRICT
AS
$BODY$
DECLARE
table_name NAME;
schema_name NAME;
is_distributed BOOL;
BEGIN
SELECT relname, nspname, replication_factor > 0
INTO STRICT table_name, schema_name, is_distributed
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
INNER JOIN _timescaledb_catalog.hypertable ht ON (ht.schema_name = n.nspname AND ht.table_name = c.relname)
WHERE c.OID = main_table;
CASE WHEN is_distributed THEN
RETURN QUERY SELECT * FROM _timescaledb_internal.hypertable_remote_size(schema_name, table_name);
ELSE
RETURN QUERY SELECT *, NULL::name FROM _timescaledb_internal.hypertable_local_size(schema_name, table_name);
END CASE;
END;
$BODY$;
--- returns total-bytes for a hypertable (includes table + index)
CREATE OR REPLACE FUNCTION hypertable_size(
main_table REGCLASS
)
RETURNS BIGINT
LANGUAGE PLPGSQL STABLE STRICT AS
$BODY$
DECLARE
num_bytes BIGINT;
BEGIN
SELECT sum(hd.total_bytes) INTO STRICT num_bytes
FROM hypertable_detailed_size(main_table) hd;
RETURN num_bytes;
END;
$BODY$;
CREATE OR REPLACE FUNCTION _timescaledb_internal.chunks_local_size(
schema_name_in name,
table_name_in name)
RETURNS TABLE (
chunk_id integer,
chunk_schema NAME,
chunk_name NAME,
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint)
LANGUAGE PLPGSQL STABLE STRICT AS
$BODY$
BEGIN
RETURN QUERY
SELECT
ch.chunk_id,
ch.chunk_schema,
ch.chunk_name,
(ch.total_bytes - COALESCE( ch.index_bytes , 0 ) - COALESCE( ch.toast_bytes, 0 ) + COALESCE( ch.compressed_heap_size , 0 ))::bigint as heap_bytes,
(COALESCE( ch.index_bytes, 0 ) + COALESCE( ch.compressed_index_size , 0) )::bigint as index_bytes,
(COALESCE( ch.toast_bytes, 0 ) + COALESCE( ch.compressed_toast_size, 0 ))::bigint as toast_bytes,
(ch.total_bytes + COALESCE( ch.compressed_heap_size, 0 ) + COALESCE( ch.compressed_index_size, 0) + COALESCE( ch.compressed_toast_size, 0 ))::bigint as total_bytes
FROM
_timescaledb_internal.hypertable_chunk_local_size ch
WHERE
ch.schema_name = schema_name_in
AND ch.table_name = table_name_in
;
END;
$BODY$;
---should return same information as chunks_local_size--
CREATE OR REPLACE FUNCTION _timescaledb_internal.chunks_remote_size(
schema_name_in name,
table_name_in name)
RETURNS TABLE (
chunk_id integer,
chunk_schema NAME,
chunk_name NAME,
table_bytes bigint,
index_bytes bigint,
toast_bytes bigint,
total_bytes bigint,
node_name NAME)
LANGUAGE PLPGSQL STABLE STRICT AS
$BODY$
BEGIN
RETURN QUERY
SELECT
entry.chunk_id,
entry.chunk_schema,
entry.chunk_name,
entry.table_bytes AS table_bytes,
entry.index_bytes AS index_bytes,
entry.toast_bytes AS toast_bytes,
entry.total_bytes AS total_bytes,
srv.node_name
FROM (
SELECT
s.node_name,
_timescaledb_internal.ping_data_node (s.node_name) AS node_up
FROM
_timescaledb_catalog.hypertable AS ht,
_timescaledb_catalog.hypertable_data_node AS s
WHERE
ht.schema_name = schema_name_in
AND ht.table_name = table_name_in
AND s.hypertable_id = ht.id
) AS srv
LEFT OUTER JOIN LATERAL _timescaledb_internal.data_node_chunk_info(
CASE WHEN srv.node_up THEN
srv.node_name
ELSE
NULL
END , schema_name_in, table_name_in) entry ON TRUE;
END;
$BODY$;
-- Get relation size of the chunks of an hypertable
-- main_table - hypertable to get size of
--
-- Returns:
-- chunk_schema - schema name for chunk
-- chunk_name - chunk table name
-- table_bytes - Disk space used by chunk table
-- index_bytes - Disk space used by indexes
-- toast_bytes - Disk space of toast tables
-- total_bytes - Disk space used in total
-- node_name - node on which chunk lives if this is
-- a distributed hypertable.
CREATE OR REPLACE FUNCTION chunks_detailed_size(
main_table REGCLASS
)
RETURNS TABLE (
chunk_schema NAME,
chunk_name NAME,
table_bytes BIGINT,
index_bytes BIGINT,
toast_bytes BIGINT,
total_bytes BIGINT,
node_name NAME)
LANGUAGE PLPGSQL STABLE STRICT
AS
$BODY$
DECLARE
table_name NAME;
schema_name NAME;
is_distributed BOOL;
BEGIN
SELECT relname, nspname, replication_factor > 0
INTO STRICT table_name, schema_name, is_distributed
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
INNER JOIN _timescaledb_catalog.hypertable ht ON (ht.schema_name = n.nspname AND ht.table_name = c.relname)
WHERE c.OID = main_table;
CASE WHEN is_distributed THEN
RETURN QUERY SELECT ch.chunk_schema, ch.chunk_name, ch.table_bytes, ch.index_bytes,
ch.toast_bytes, ch.total_bytes, ch.node_name
FROM _timescaledb_internal.chunks_remote_size(schema_name, table_name) ch;
ELSE
RETURN QUERY SELECT chl.chunk_schema, chl.chunk_name, chl.table_bytes, chl.index_bytes,
chl.toast_bytes, chl.total_bytes, NULL::NAME
FROM _timescaledb_internal.chunks_local_size(schema_name, table_name) chl;
END CASE;
END;
$BODY$;
-- Get sizes of indexes on a hypertable
--
-- main_table - hypertable to get index sizes of
--
-- Returns:
-- index_name - index on hyper table
-- total_bytes - size of index on disk
CREATE OR REPLACE FUNCTION indexes_relation_size(
main_table REGCLASS
)
RETURNS TABLE (index_name TEXT,
total_bytes BIGINT)
LANGUAGE PLPGSQL STABLE STRICT
AS
$BODY$
<<main>>
DECLARE
table_name NAME;
schema_name NAME;
BEGIN
SELECT relname, nspname
INTO STRICT table_name, schema_name
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
WHERE c.OID = main_table;
RETURN QUERY
SELECT format('%I.%I', h.schema_name, ci.hypertable_index_name),
sum(pg_relation_size(c.oid))::bigint
FROM
pg_class c,
pg_namespace n,
_timescaledb_catalog.hypertable h,
_timescaledb_catalog.chunk ch,
_timescaledb_catalog.chunk_index ci
WHERE ch.schema_name = n.nspname
AND c.relnamespace = n.oid
AND c.relname = ci.index_name
AND ch.id = ci.chunk_id
AND h.id = ci.hypertable_id
AND h.schema_name = main.schema_name
AND h.table_name = main.table_name
GROUP BY h.schema_name, ci.hypertable_index_name;
END;
$BODY$;
------- REMOVE chunk_relation_size -----
-- Get relation size of the chunks of an hypertable
-- like pg_relation_size
-- (https://www.postgresql.org/docs/9.6/static/functions-admin.html#FUNCTIONS-ADMIN-DBSIZE)
--
-- main_table - hypertable to get size of
--
-- Returns:
-- chunk_id - Timescaledb id of a chunk
-- chunk_table - Table used for the chunk
-- partitioning_columns - Partitioning column names
-- partitioning_column_types - Type of partitioning columns
-- partitioning_hash_functions - Hash functions of partitioning columns
-- ranges - Partition ranges for each dimension of the chunk
-- table_bytes - Disk space used by main_table
-- index_bytes - Disk space used by indexes
-- toast_bytes - Disk space of toast tables
-- total_bytes - Disk space used in total
CREATE OR REPLACE FUNCTION chunk_relation_size(
main_table REGCLASS
)
RETURNS TABLE (chunk_id INT,
chunk_table TEXT,
partitioning_columns NAME[],
partitioning_column_types REGTYPE[],
partitioning_hash_functions TEXT[],
ranges int8range[],
table_bytes BIGINT,
index_bytes BIGINT,
toast_bytes BIGINT,
total_bytes BIGINT)
LANGUAGE PLPGSQL STABLE STRICT
AS
$BODY$
DECLARE
table_name NAME;
schema_name NAME;
BEGIN
SELECT relname, nspname
INTO STRICT table_name, schema_name
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
WHERE c.OID = main_table;
RETURN QUERY EXECUTE format(
$$
SELECT chunk_id,
chunk_table,
partitioning_columns,
partitioning_column_types,
partitioning_hash_functions,
ranges,
table_bytes,
index_bytes,
toast_bytes,
total_bytes
FROM (
SELECT *,
total_bytes-index_bytes-COALESCE(toast_bytes,0) AS table_bytes
FROM (
SELECT c.id as chunk_id,
format('%%I.%%I', c.schema_name, c.table_name) as chunk_table,
pg_total_relation_size(format('%%I.%%I', c.schema_name, c.table_name)) AS total_bytes,
pg_indexes_size(format('%%I.%%I', c.schema_name, c.table_name)) AS index_bytes,
pg_total_relation_size(reltoastrelid) AS toast_bytes,
array_agg(d.column_name ORDER BY d.interval_length, d.column_name ASC) as partitioning_columns,
array_agg(d.column_type ORDER BY d.interval_length, d.column_name ASC) as partitioning_column_types,
array_agg(d.partitioning_func_schema || '.' || d.partitioning_func ORDER BY d.interval_length, d.column_name ASC) as partitioning_hash_functions,
array_agg(int8range(range_start, range_end) ORDER BY d.interval_length, d.column_name ASC) as ranges
FROM
_timescaledb_catalog.hypertable h,
_timescaledb_catalog.chunk c,
_timescaledb_catalog.chunk_constraint cc,
_timescaledb_catalog.dimension d,
_timescaledb_catalog.dimension_slice ds,
pg_class pgc,
pg_namespace pns
WHERE h.schema_name = %L
AND h.table_name = %L
AND pgc.relname = c.table_name
AND pns.oid = pgc.relnamespace
AND pns.nspname = c.schema_name
AND relkind = 'r'
AND c.hypertable_id = h.id
AND c.id = cc.chunk_id
AND cc.dimension_slice_id = ds.id
AND ds.dimension_id = d.id
GROUP BY c.id, pgc.reltoastrelid, pgc.oid ORDER BY c.id
) sub1
) sub2;
$$,
schema_name, table_name);
END;
$BODY$;
-- Get relation size of the chunks of an hypertable
-- like pg_relation_size
-- (https://www.postgresql.org/docs/9.6/static/functions-admin.html#FUNCTIONS-ADMIN-DBSIZE)
--
-- main_table - hypertable to get size of
--
-- Returns:
-- chunk_id - Timescaledb id of a chunk
-- chunk_table - Table used for the chunk
-- partitioning_columns - Partitioning column names
-- partitioning_column_types - Type of partitioning columns
-- partitioning_hash_functions - Hash functions of partitioning columns
-- ranges - Partition ranges for each dimension of the chunk
-- table_size - Pretty output of table_bytes
-- index_size - Pretty output of index_bytes
-- toast_size - Pretty output of toast_bytes
-- total_size - Pretty output of total_bytes
CREATE OR REPLACE FUNCTION chunk_relation_size_pretty(
main_table REGCLASS
)
RETURNS TABLE (chunk_id INT,
chunk_table TEXT,
partitioning_columns NAME[],
partitioning_column_types REGTYPE[],
partitioning_hash_functions TEXT[],
ranges TEXT[],
table_size TEXT,
index_size TEXT,
toast_size TEXT,
total_size TEXT
)
LANGUAGE PLPGSQL STABLE STRICT
AS
$BODY$
DECLARE
table_name NAME;
schema_name NAME;
BEGIN
SELECT relname, nspname
INTO STRICT table_name, schema_name
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
WHERE c.OID = main_table;
RETURN QUERY EXECUTE format(
$$
SELECT chunk_id,
chunk_table,
partitioning_columns,
partitioning_column_types,
partitioning_functions,
ranges,
pg_size_pretty(table_bytes) AS table,
pg_size_pretty(index_bytes) AS index,
pg_size_pretty(toast_bytes) AS toast,
pg_size_pretty(total_bytes) AS total
FROM (
SELECT *,
total_bytes-index_bytes-COALESCE(toast_bytes,0) AS table_bytes
FROM (
SELECT c.id as chunk_id,
format('%%I.%%I', c.schema_name, c.table_name) as chunk_table,
pg_total_relation_size(format('%%I.%%I', c.schema_name, c.table_name)) AS total_bytes,
pg_indexes_size(format('%%I.%%I', c.schema_name, c.table_name)) AS index_bytes,
pg_total_relation_size(reltoastrelid) AS toast_bytes,
array_agg(d.column_name ORDER BY d.interval_length, d.column_name ASC) as partitioning_columns,
array_agg(d.column_type ORDER BY d.interval_length, d.column_name ASC) as partitioning_column_types,
array_agg(d.partitioning_func_schema || '.' || d.partitioning_func ORDER BY d.interval_length, d.column_name ASC) as partitioning_functions,
array_agg('[' || _timescaledb_internal.range_value_to_pretty(range_start, column_type) ||
',' ||
_timescaledb_internal.range_value_to_pretty(range_end, column_type) || ')' ORDER BY d.interval_length, d.column_name ASC) as ranges
FROM
_timescaledb_catalog.hypertable h,
_timescaledb_catalog.chunk c,
_timescaledb_catalog.chunk_constraint cc,
_timescaledb_catalog.dimension d,
_timescaledb_catalog.dimension_slice ds,
pg_class pgc,
pg_namespace pns
WHERE h.schema_name = %L
AND h.table_name = %L
AND pgc.relname = c.table_name
AND pns.oid = pgc.relnamespace
AND pns.nspname = c.schema_name
AND relkind = 'r'
AND c.hypertable_id = h.id
AND c.id = cc.chunk_id
AND cc.dimension_slice_id = ds.id
AND ds.dimension_id = d.id
GROUP BY c.id, pgc.reltoastrelid, pgc.oid ORDER BY c.id
) sub1
) sub2;
$$,
schema_name, table_name);
END;
$BODY$;
CREATE OR REPLACE FUNCTION _timescaledb_internal.range_value_to_pretty(
time_value BIGINT,
column_type REGTYPE
)
RETURNS TEXT LANGUAGE PLPGSQL STABLE AS
$BODY$
DECLARE
BEGIN
IF NOT _timescaledb_internal.dimension_is_finite(time_value) THEN
RETURN '';
END IF;
IF time_value IS NULL THEN
RETURN format('%L', NULL);
END IF;
CASE column_type
WHEN 'BIGINT'::regtype, 'INTEGER'::regtype, 'SMALLINT'::regtype THEN
RETURN format('%L', time_value); -- scale determined by user.
WHEN 'TIMESTAMP'::regtype, 'TIMESTAMPTZ'::regtype THEN
-- assume time_value is in microsec
RETURN format('%1$L', _timescaledb_internal.to_timestamp(time_value)); -- microseconds
WHEN 'DATE'::regtype THEN
RETURN format('%L', timezone('UTC',_timescaledb_internal.to_timestamp(time_value))::date);
ELSE
RETURN time_value;
END CASE;
END
$BODY$;
-- Convenience function to return approximate row count
--
-- main_table - hypertable to get approximate row count for; if NULL, get count
-- for all hypertables
--
-- Returns:
-- schema_name - Schema name of the hypertable
-- table_name - Table name of the hypertable
-- row_estimate - Estimated number of rows according to catalog tables
CREATE OR REPLACE FUNCTION hypertable_approximate_row_count(
main_table REGCLASS = NULL
)
RETURNS TABLE (schema_name NAME,
table_name NAME,
row_estimate BIGINT
) LANGUAGE PLPGSQL VOLATILE
AS
$BODY$
<<main>>
DECLARE
table_name NAME;
schema_name NAME;
BEGIN
IF main_table IS NOT NULL THEN
SELECT relname, nspname
INTO STRICT table_name, schema_name
FROM pg_class c
INNER JOIN pg_namespace n ON (n.OID = c.relnamespace)
WHERE c.OID = main_table;
END IF;
-- Thanks to @fvannee on Github for providing the initial draft
-- of this query
RETURN QUERY
SELECT h.schema_name,
h.table_name,
row_estimate.row_estimate
FROM _timescaledb_catalog.hypertable h
CROSS JOIN LATERAL (
SELECT sum(cl.reltuples)::BIGINT AS row_estimate
FROM _timescaledb_catalog.chunk c
JOIN pg_class cl ON cl.relname = c.table_name
WHERE c.hypertable_id = h.id
GROUP BY h.schema_name, h.table_name
) row_estimate
WHERE (main.table_name IS NULL OR h.table_name = main.table_name)
AND (main.schema_name IS NULL OR h.schema_name = main.schema_name)
ORDER BY h.schema_name, h.table_name;
END
$BODY$;