collectd_insert.sql

-- Description:
-- ------------
--
-- This is an example schema for storing collectd metrics in a TimescaleDB
-- dabase (see https://www.timescale.com/).  It is based on the PostgreSQL
-- schema contributed by Sebastian Harl.
--
-- Prerequisities:
-- ---------------
--
-- You must have an existing database with the timescaledb extension
-- installed.  For example:
--
--   postgres=# create database collectd;
--   postgres=# \c collectd
--   collectd=# create extension timescaledb;
--
-- Sample configuration:
-- ---------------------
--
-- <Plugin postgresql>
--     <Writer sqlstore>
--         Statement "SELECT collectd_insert($1, $2, $3, $4, $5, $6, $7, $8, $9);"
--     </Writer>
--     <Database collectd>
--         # ...
--         Writer sqlstore
--     </Database>
-- </Plugin>
--
-- Sample query:
-- -------------
--
-- This is an example query for use in Grafana.
--
--   SELECT
--     time_bucket($__interval, tstamp) as time,
--     type_inst as metric,
--     avg(value)
--   FROM
--     collectd
--   WHERE
--     tstamp >= $__timeFrom() and tstamp <= $__timeTo() and
--     host = 'stats.house' and
--     plugin = 'memory'
--     group by time, type_inst
--     order by time asc
--
-- This would get translated into the following SQL:
--
--   SELECT
--     time_bucket('1m', tstamp) as time,
--     type_inst as metric,
--     avg(value)
--   FROM
--     collectd
--   WHERE
--     tstamp >= to_timestamp(1512444891) and tstamp <= to_timestamp(1512448491) and
--     host = 'stats.house' and
--     plugin = 'memory'
--     group by time, type_inst
--     order by time asc

CREATE TABLE identifiers (
    id integer NOT NULL PRIMARY KEY,
    host character varying(64) NOT NULL,
    plugin character varying(64) NOT NULL,
    plugin_inst character varying(64) DEFAULT NULL::character varying,
    type character varying(64) NOT NULL,
    type_inst character varying(64) DEFAULT NULL::character varying,

    UNIQUE (host, plugin, plugin_inst, type, type_inst)
);

CREATE SEQUENCE identifiers_id_seq
    OWNED BY identifiers.id
    START WITH 1
    INCREMENT BY 1
    NO MINVALUE
    NO MAXVALUE
    CACHE 1;

ALTER TABLE identifiers
    ALTER COLUMN id
    SET DEFAULT nextval('identifiers_id_seq'::regclass);

-- create indexes for the identifier fields
CREATE INDEX identifiers_host ON identifiers USING btree (host);
CREATE INDEX identifiers_plugin ON identifiers USING btree (plugin);
CREATE INDEX identifiers_plugin_inst ON identifiers USING btree (plugin_inst);
CREATE INDEX identifiers_type ON identifiers USING btree (type);
CREATE INDEX identifiers_type_inst ON identifiers USING btree (type_inst);

CREATE TABLE "values" (
    id integer NOT NULL
        REFERENCES identifiers
        ON DELETE cascade,
    tstamp timestamp with time zone NOT NULL,
    name character varying(64) NOT NULL,
    value double precision NOT NULL,

    UNIQUE(tstamp, id, name)
);

-- This demonstrates how to set a custom partition for your
-- hypertable. The default partition is 30 days.  See
-- http://docs.timescale.com/latest/api for additional information.
SELECT create_hypertable('values', 'tstamp',
	chunk_time_interval => interval '1 day');

CREATE OR REPLACE VIEW collectd
    AS SELECT host, plugin, plugin_inst, type, type_inst,
            host
                || '/' || plugin
                || CASE
                    WHEN plugin_inst IS NOT NULL THEN '-'
                    ELSE ''
                END
                || coalesce(plugin_inst, '')
                || '/' || type
                || CASE
                    WHEN type_inst IS NOT NULL THEN '-'
                    ELSE ''
                END
                || coalesce(type_inst, '') AS identifier,
            tstamp, name, value
        FROM identifiers JOIN values ON values.id = identifiers.id;

CREATE OR REPLACE FUNCTION collectd_insert(
        timestamp with time zone, character varying,
        character varying, character varying,
        character varying, character varying,
        character varying[], character varying[], double precision[]
    ) RETURNS void
    LANGUAGE plpgsql
    AS $_$
DECLARE
    p_time alias for $1;
    p_host alias for $2;
    p_plugin alias for $3;
    p_plugin_instance alias for $4;
    p_type alias for $5;
    p_type_instance alias for $6;
    p_value_names alias for $7;
    -- don't use the type info; for 'StoreRates true' it's 'gauge' anyway
    -- p_type_names alias for $8;
    p_values alias for $9;
    ds_id integer;
    i integer;
BEGIN
    SELECT id INTO ds_id
        FROM identifiers
        WHERE host = p_host
            AND plugin = p_plugin
            AND COALESCE(plugin_inst, '') = COALESCE(p_plugin_instance, '')
            AND type = p_type
            AND COALESCE(type_inst, '') = COALESCE(p_type_instance, '');
    IF NOT FOUND THEN
        INSERT INTO identifiers (host, plugin, plugin_inst, type, type_inst)
            VALUES (p_host, p_plugin, p_plugin_instance, p_type, p_type_instance)
            RETURNING id INTO ds_id;
    END IF;
    i := 1;
    LOOP
        EXIT WHEN i > array_upper(p_value_names, 1);
        INSERT INTO values (id, tstamp, name, value)
            VALUES (ds_id, p_time, p_value_names[i], p_values[i]);
        i := i + 1;
    END LOOP;
END;
$_$;

query.results

                                                                              QUERY PLAN                                                                               
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Sort  (cost=40423.86..40426.36 rows=1000 width=25) (actual time=1824.830..1827.008 rows=6804 loops=1)
   Sort Key: (time_bucket('00:01:00'::interval, "values".tstamp))
   Sort Method: quicksort  Memory: 724kB
   ->  HashAggregate  (cost=40359.03..40374.03 rows=1000 width=25) (actual time=1815.610..1820.219 rows=6804 loops=1)
         Group Key: time_bucket('00:01:00'::interval, "values".tstamp), identifiers.type_inst
         ->  Hash Join  (cost=4.35..40020.48 rows=45140 width=25) (actual time=0.490..1764.443 rows=40824 loops=1)
               Hash Cond: ("values".id = identifiers.id)
               ->  Append  (cost=0.00..33809.32 rows=1504682 width=20) (actual time=0.049..1388.556 rows=1503682 loops=1)
                     ->  Seq Scan on "values"  (cost=0.00..0.00 rows=1 width=20) (actual time=0.009..0.009 rows=0 loops=1)
                           Filter: ((tstamp >= '2017-12-04 21:22:53+00'::timestamp with time zone) AND (tstamp <= '2017-12-05 21:22:53+00'::timestamp with time zone))
                     ->  Seq Scan on _hyper_1_1_chunk  (cost=0.00..33809.32 rows=1504681 width=20) (actual time=0.039..997.507 rows=1503682 loops=1)
                           Filter: ((tstamp >= '2017-12-04 21:22:53+00'::timestamp with time zone) AND (tstamp <= '2017-12-05 21:22:53+00'::timestamp with time zone))
                           Rows Removed by Filter: 1079
               ->  Hash  (cost=4.28..4.28 rows=6 width=13) (actual time=0.134..0.134 rows=6 loops=1)
                     Buckets: 1024  Batches: 1  Memory Usage: 9kB
                     ->  Seq Scan on identifiers  (cost=0.00..4.28 rows=6 width=13) (actual time=0.039..0.113 rows=6 loops=1)
                           Filter: (((host)::text = 'stats.house'::text) AND ((plugin)::text = 'memory'::text))
                           Rows Removed by Filter: 146
 Planning time: 4.237 ms
 Execution time: 1828.628 ms
(20 rows)

query.sql

explain analyze SELECT
  time_bucket('1m', tstamp) as time,
  type_inst as metric,
  avg(value)
FROM
  collectd
WHERE
  tstamp >= to_timestamp(1512422573) and tstamp <= to_timestamp(1512508973) and
  host = 'stats.house' and
  plugin = 'memory'
  group by time, type_inst
  order by time asc

Hi there,

thank you for this, makes my life much easier.
However, I don't really like that "name" is not normalized. Is there a reason for that? I'd guess a plugin always sends the same data.

Maybe it could be moved into the identifiers table? What do you think?

Thank you in advance!
Bests,
Semir

Hi, I'm glad this was helpful! It's been four years since I put this together and I'm honestly not sure what I was thinking at the time :). I'm no longer using timescaledb as a metrics backend for anything, so I'm not going to be able to provide useful answers to most questions.

Hi,
thank you for your quick response.
I don't think that I would have any more questions :D I think it ought to work.

Thanks anyways :)
Take care!

Since this is one of the top most search results for "collectd" "timescaledb", one minor comment from my side.
character varying(64) makes the df plugin unhappy. In the age of containers mount paths can be super long and exceed 64 characters easy.