Erik Nordström ece582d458 Add mappings table for remote hypertables
In a multi-node (clustering) setup, TimescaleDB needs to track which
remote servers have data for a particular distributed hypertable. It
also needs to know which servers to place new chunks on and to use in
queries against a distributed hypertable.

A new metadata table, `hypertable_server` is added to map a local
hypertable ID to a hypertable ID on a remote server. We require that
the remote hypertable has the same schema and name as the local
hypertable.

When a local server is removed (using `DROP SERVER` or our
`delete_server()`), all remote hypertable mappings for that server
should also be removed.
2020-05-27 17:31:09 +02:00

377 lines
19 KiB
SQL

-- 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.
--NOTICE: UPGRADE-SCRIPT-NEEDED contents in this file are not auto-upgraded.
-- This file contains table definitions for various abstractions and data
-- structures for representing hypertables and lower-level concepts.
-- Hypertable
-- ==========
--
-- The hypertable is an abstraction that represents a table that is
-- partitioned in N dimensions, where each dimension maps to a column
-- in the table. A dimension can either be 'open' or 'closed', which
-- reflects the scheme that divides the dimension's keyspace into
-- "slices".
--
-- Conceptually, a partition -- called a "chunk", is a hypercube in
-- the N-dimensional space. A chunk stores a subset of the
-- hypertable's tuples on disk in its own distinct table. The slices
-- that span the chunk's hypercube each correspond to a constraint on
-- the chunk's table, enabling constraint exclusion during queries on
-- the hypertable's data.
--
--
-- Open dimensions
------------------
-- An open dimension does on-demand slicing, creating a new slice
-- based on a configurable interval whenever a tuple falls outside the
-- existing slices. Open dimensions fit well with columns that are
-- incrementally increasing, such as time-based ones.
--
-- Closed dimensions
--------------------
-- A closed dimension completely divides its keyspace into a
-- configurable number of slices. The number of slices can be
-- reconfigured, but the new partitioning only affects newly created
-- chunks.
--
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.hypertable (
id SERIAL PRIMARY KEY,
schema_name NAME NOT NULL CHECK (schema_name != '_timescaledb_catalog'),
table_name NAME NOT NULL,
associated_schema_name NAME NOT NULL,
associated_table_prefix NAME NOT NULL,
num_dimensions SMALLINT NOT NULL,
chunk_sizing_func_schema NAME NOT NULL,
chunk_sizing_func_name NAME NOT NULL,
chunk_target_size BIGINT NOT NULL CHECK (chunk_target_size >= 0), -- size in bytes
compressed BOOLEAN NOT NULL DEFAULT false,
compressed_hypertable_id INTEGER REFERENCES _timescaledb_catalog.hypertable(id),
replication_factor SMALLINT NULL CHECK (replication_factor > 0),
UNIQUE (id, schema_name),
UNIQUE (schema_name, table_name),
UNIQUE (associated_schema_name, associated_table_prefix),
constraint hypertable_dim_compress_check check ( num_dimensions > 0 or compressed = true ),
constraint hypertable_compress_check check ( compressed = false or (compressed = true and compressed_hypertable_id is null ))
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.hypertable', '');
SELECT pg_catalog.pg_extension_config_dump(pg_get_serial_sequence('_timescaledb_catalog.hypertable','id'), '');
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.hypertable_server (
hypertable_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.hypertable(id),
server_hypertable_id INTEGER NULL,
server_name NAME NOT NULL,
UNIQUE (server_hypertable_id, server_name),
UNIQUE (hypertable_id, server_name)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.hypertable_server', '');
-- The tablespace table maps tablespaces to hypertables.
-- This allows spreading a hypertable's chunks across multiple disks.
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.tablespace (
id SERIAL PRIMARY KEY,
hypertable_id INT NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
tablespace_name NAME NOT NULL,
UNIQUE (hypertable_id, tablespace_name)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.tablespace', '');
-- A dimension represents an axis along which data is partitioned.
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.dimension (
id SERIAL NOT NULL PRIMARY KEY,
hypertable_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
column_name NAME NOT NULL,
column_type REGTYPE NOT NULL,
aligned BOOLEAN NOT NULL,
-- closed dimensions
num_slices SMALLINT NULL,
partitioning_func_schema NAME NULL,
partitioning_func NAME NULL,
-- open dimensions (e.g., time)
interval_length BIGINT NULL CHECK(interval_length IS NULL OR interval_length > 0),
integer_now_func_schema NAME NULL,
integer_now_func NAME NULL,
CHECK (
(partitioning_func_schema IS NULL AND partitioning_func IS NULL) OR
(partitioning_func_schema IS NOT NULL AND partitioning_func IS NOT NULL)
),
CHECK (
(num_slices IS NULL AND interval_length IS NOT NULL) OR
(num_slices IS NOT NULL AND interval_length IS NULL)
),
CHECK (
(integer_now_func_schema IS NULL AND integer_now_func IS NULL) OR
(integer_now_func_schema IS NOT NULL AND integer_now_func IS NOT NULL)
),
UNIQUE (hypertable_id, column_name)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.dimension', '');
SELECT pg_catalog.pg_extension_config_dump(pg_get_serial_sequence('_timescaledb_catalog.dimension','id'), '');
-- A dimension slice defines a keyspace range along a dimension axis.
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.dimension_slice (
id SERIAL NOT NULL PRIMARY KEY,
dimension_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.dimension(id) ON DELETE CASCADE,
range_start BIGINT NOT NULL,
range_end BIGINT NOT NULL,
CHECK (range_start <= range_end),
UNIQUE (dimension_id, range_start, range_end)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.dimension_slice', '');
SELECT pg_catalog.pg_extension_config_dump(pg_get_serial_sequence('_timescaledb_catalog.dimension_slice','id'), '');
-- A chunk is a partition (hypercube) in an N-dimensional
-- hyperspace. Each chunk is associated with N constraints that define
-- the chunk's hypercube. Tuples that fall within the chunk's
-- hypercube are stored in the chunk's data table, as given by
-- 'schema_name' and 'table_name'.
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.chunk (
id SERIAL NOT NULL PRIMARY KEY,
hypertable_id INT NOT NULL REFERENCES _timescaledb_catalog.hypertable(id),
schema_name NAME NOT NULL,
table_name NAME NOT NULL,
compressed_chunk_id INTEGER REFERENCES _timescaledb_catalog.chunk(id),
dropped BOOLEAN NOT NULL DEFAULT false,
UNIQUE (schema_name, table_name)
);
CREATE INDEX IF NOT EXISTS chunk_hypertable_id_idx
ON _timescaledb_catalog.chunk(hypertable_id);
CREATE INDEX IF NOT EXISTS chunk_compressed_chunk_id_idx
ON _timescaledb_catalog.chunk(compressed_chunk_id);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.chunk', '');
SELECT pg_catalog.pg_extension_config_dump(pg_get_serial_sequence('_timescaledb_catalog.chunk','id'), '');
-- A chunk constraint maps a dimension slice to a chunk. Each
-- constraint associated with a chunk will also be a table constraint
-- on the chunk's data table.
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.chunk_constraint (
chunk_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.chunk(id),
dimension_slice_id INTEGER NULL REFERENCES _timescaledb_catalog.dimension_slice(id),
constraint_name NAME NOT NULL,
hypertable_constraint_name NAME NULL,
UNIQUE(chunk_id, constraint_name)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.chunk_constraint', '');
CREATE INDEX IF NOT EXISTS chunk_constraint_chunk_id_dimension_slice_id_idx
ON _timescaledb_catalog.chunk_constraint(chunk_id, dimension_slice_id);
CREATE SEQUENCE IF NOT EXISTS _timescaledb_catalog.chunk_constraint_name;
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.chunk_constraint_name', '');
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.chunk_index (
chunk_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.chunk(id) ON DELETE CASCADE,
index_name NAME NOT NULL,
hypertable_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
hypertable_index_name NAME NOT NULL,
UNIQUE(chunk_id, index_name)
);
CREATE INDEX IF NOT EXISTS chunk_index_hypertable_id_hypertable_index_name_idx
ON _timescaledb_catalog.chunk_index(hypertable_id, hypertable_index_name);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.chunk_index', '');
-- Default jobs are given the id space [1,1000). User-installed jobs and any jobs created inside tests
-- are given the id space [1000, INT_MAX). That way, we do not pg_dump jobs that are always default-installed
-- inside other .sql scripts. This avoids insertion conflicts during pg_restore.
CREATE SEQUENCE IF NOT EXISTS _timescaledb_config.bgw_job_id_seq MINVALUE 1000;
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_config.bgw_job_id_seq', '');
CREATE TABLE IF NOT EXISTS _timescaledb_config.bgw_job (
id INTEGER PRIMARY KEY DEFAULT nextval('_timescaledb_config.bgw_job_id_seq'),
application_name NAME NOT NULL,
job_type NAME NOT NULL,
schedule_interval INTERVAL NOT NULL,
max_runtime INTERVAL NOT NULL,
max_retries INT NOT NULL,
retry_period INTERVAL NOT NULL,
CONSTRAINT valid_job_type CHECK (job_type IN ('telemetry_and_version_check_if_enabled', 'reorder', 'drop_chunks', 'continuous_aggregate', 'compress_chunks'))
);
ALTER SEQUENCE _timescaledb_config.bgw_job_id_seq OWNED BY _timescaledb_config.bgw_job.id;
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_config.bgw_job', 'WHERE id >= 1000');
CREATE TABLE IF NOT EXISTS _timescaledb_internal.bgw_job_stat (
job_id INTEGER PRIMARY KEY REFERENCES _timescaledb_config.bgw_job(id) ON DELETE CASCADE,
last_start TIMESTAMPTZ NOT NULL DEFAULT NOW(),
last_finish TIMESTAMPTZ NOT NULL,
next_start TIMESTAMPTZ NOT NULL,
last_successful_finish TIMESTAMPTZ NOT NULL,
last_run_success BOOL NOT NULL,
total_runs BIGINT NOT NULL,
total_duration INTERVAL NOT NULL,
total_successes BIGINT NOT NULL,
total_failures BIGINT NOT NULL,
total_crashes BIGINT NOT NULL,
consecutive_failures INT NOT NULL,
consecutive_crashes INT NOT NULL
);
--The job_stat table is not dumped by pg_dump on purpose because
--the statistics probably aren't very meaningful across instances.
--Now we define the argument tables for available BGW policies.
CREATE TABLE IF NOT EXISTS _timescaledb_config.bgw_policy_reorder (
job_id INTEGER PRIMARY KEY REFERENCES _timescaledb_config.bgw_job(id) ON DELETE CASCADE,
hypertable_id INTEGER UNIQUE NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
hypertable_index_name NAME NOT NULL
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_config.bgw_policy_reorder', '');
CREATE TABLE IF NOT EXISTS _timescaledb_config.bgw_policy_drop_chunks (
job_id INTEGER PRIMARY KEY REFERENCES _timescaledb_config.bgw_job(id) ON DELETE CASCADE,
hypertable_id INTEGER UNIQUE NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
older_than _timescaledb_catalog.ts_interval NOT NULL,
cascade BOOLEAN NOT NULL,
cascade_to_materializations BOOLEAN,
CONSTRAINT valid_older_than CHECK(_timescaledb_internal.valid_ts_interval(older_than))
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_config.bgw_policy_drop_chunks', '');
----- End BGW policy table definitions
-- Now we define a special stats table for each job/chunk pair. This will be used by the scheduler
-- to determine whether to run a specific job on a specific chunk.
CREATE TABLE IF NOT EXISTS _timescaledb_internal.bgw_policy_chunk_stats (
job_id INTEGER NOT NULL REFERENCES _timescaledb_config.bgw_job(id) ON DELETE CASCADE,
chunk_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.chunk(id) ON DELETE CASCADE,
num_times_job_run INTEGER,
last_time_job_run TIMESTAMPTZ,
UNIQUE(job_id,chunk_id)
);
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.metadata (
key NAME NOT NULL PRIMARY KEY,
value TEXT NOT NULL,
include_in_telemetry BOOLEAN NOT NULL
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.metadata', $$WHERE key='exported_uuid'$$);
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.continuous_agg (
mat_hypertable_id INTEGER PRIMARY KEY REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
raw_hypertable_id INTEGER NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
user_view_schema NAME NOT NULL,
user_view_name NAME NOT NULL,
partial_view_schema NAME NOT NULL,
partial_view_name NAME NOT NULL,
bucket_width BIGINT NOT NULL,
job_id INTEGER UNIQUE NOT NULL REFERENCES _timescaledb_config.bgw_job(id) ON DELETE RESTRICT,
refresh_lag BIGINT NOT NULL,
direct_view_schema NAME NOT NULL,
direct_view_name NAME NOT NULL,
max_interval_per_job BIGINT NOT NULL,
ignore_invalidation_older_than BIGINT NOT NULL DEFAULT BIGINT '9223372036854775807',
materialized_only BOOL NOT NULL DEFAULT false,
UNIQUE(user_view_schema, user_view_name),
UNIQUE(partial_view_schema, partial_view_name)
);
CREATE INDEX IF NOT EXISTS continuous_agg_raw_hypertable_id_idx
ON _timescaledb_catalog.continuous_agg(raw_hypertable_id);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_agg', '');
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.continuous_aggs_invalidation_threshold(
hypertable_id INTEGER PRIMARY KEY REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
watermark BIGINT NOT NULL
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_aggs_invalidation_threshold', '');
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.continuous_aggs_completed_threshold(
materialization_id INTEGER PRIMARY KEY
REFERENCES _timescaledb_catalog.continuous_agg(mat_hypertable_id)
ON DELETE CASCADE,
watermark BIGINT NOT NULL --exclusive (everything up to but not including watermark is done)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_aggs_completed_threshold', '');
-- this does not have an FK on the materialization table since INSERTs to this
-- table are performance critical
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.continuous_aggs_hypertable_invalidation_log(
hypertable_id INTEGER NOT NULL,
modification_time BIGINT NOT NULL, --now time for txn when the raw table was modified
lowest_modified_value BIGINT NOT NULL,
greatest_modified_value BIGINT NOT NULL
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_aggs_hypertable_invalidation_log', '');
CREATE INDEX continuous_aggs_hypertable_invalidation_log_idx
ON _timescaledb_catalog.continuous_aggs_hypertable_invalidation_log (hypertable_id, lowest_modified_value ASC);
-- per cagg copy of invalidation log
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.continuous_aggs_materialization_invalidation_log(
materialization_id INTEGER
REFERENCES _timescaledb_catalog.continuous_agg(mat_hypertable_id)
ON DELETE CASCADE,
modification_time BIGINT NOT NULL, --now time for txn when the raw table was modified
lowest_modified_value BIGINT NOT NULL,
greatest_modified_value BIGINT NOT NULL
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.continuous_aggs_materialization_invalidation_log', '');
CREATE INDEX continuous_aggs_materialization_invalidation_log_idx
ON _timescaledb_catalog.continuous_aggs_materialization_invalidation_log (materialization_id, lowest_modified_value ASC);
/* the source of this data is the enum from the source code that lists
* the algorithms. This table is NOT dumped.
*/
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.compression_algorithm(
id SMALLINT PRIMARY KEY,
version SMALLINT NOT NULL,
name NAME NOT NULL,
description TEXT
);
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.hypertable_compression (
hypertable_id INTEGER REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
attname NAME NOT NULL,
compression_algorithm_id SMALLINT REFERENCES _timescaledb_catalog.compression_algorithm(id),
segmentby_column_index SMALLINT ,
orderby_column_index SMALLINT,
orderby_asc BOOLEAN,
orderby_nullsfirst BOOLEAN,
PRIMARY KEY (hypertable_id, attname),
UNIQUE (hypertable_id, segmentby_column_index),
UNIQUE (hypertable_id, orderby_column_index)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.hypertable_compression', '');
CREATE TABLE IF NOT EXISTS _timescaledb_catalog.compression_chunk_size (
chunk_id INTEGER REFERENCES _timescaledb_catalog.chunk(id) ON DELETE CASCADE,
compressed_chunk_id INTEGER REFERENCES _timescaledb_catalog.chunk(id) ON DELETE CASCADE,
uncompressed_heap_size BIGINT NOT NULL,
uncompressed_toast_size BIGINT NOT NULL,
uncompressed_index_size BIGINT NOT NULL,
compressed_heap_size BIGINT NOT NULL,
compressed_toast_size BIGINT NOT NULL,
compressed_index_size BIGINT NOT NULL,
PRIMARY KEY(chunk_id, compressed_chunk_id)
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_catalog.compression_chunk_size', '');
CREATE TABLE IF NOT EXISTS _timescaledb_config.bgw_policy_compress_chunks(
job_id INTEGER PRIMARY KEY REFERENCES _timescaledb_config.bgw_job(id) ON DELETE CASCADE,
hypertable_id INTEGER UNIQUE NOT NULL REFERENCES _timescaledb_catalog.hypertable(id) ON DELETE CASCADE,
older_than _timescaledb_catalog.ts_interval NOT NULL,
CONSTRAINT valid_older_than CHECK(_timescaledb_internal.valid_ts_interval(older_than))
);
SELECT pg_catalog.pg_extension_config_dump('_timescaledb_config.bgw_policy_compress_chunks', '');
-- Set table permissions
-- We need to grant SELECT to PUBLIC for all tables even those not
-- marked as being dumped because pg_dump will try to access all
-- tables initially to detect inheritance chains and then decide
-- which objects actually need to be dumped.
GRANT SELECT ON ALL TABLES IN SCHEMA _timescaledb_catalog TO PUBLIC;
GRANT SELECT ON ALL TABLES IN SCHEMA _timescaledb_config TO PUBLIC;
GRANT SELECT ON ALL TABLES IN SCHEMA _timescaledb_internal TO PUBLIC;
GRANT SELECT ON ALL SEQUENCES IN SCHEMA _timescaledb_catalog TO PUBLIC;
GRANT SELECT ON ALL SEQUENCES IN SCHEMA _timescaledb_config TO PUBLIC;
GRANT SELECT ON ALL SEQUENCES IN SCHEMA _timescaledb_internal TO PUBLIC;