After I successfully migrated Zabbix from MySQL to PostgreSQL in the previous article How to migrate Zabbix from MySQL to PostgreSQL with minimal downtime , it became necessary to take the next step - to migrate the database to TimescaleDB, since for her sake everything was started.
The reader may wonder: why is this article needed if there is a simple and understandable manual ?
But the problem, as in the previous article, is hidden in the downtime. The manual clearly states:
The migration of existing history and trend data may take a lot of time. Zabbix server and frontend must be down for the period of migration.
:
- Zabbix 5.0
- PostgreSQL 12
- ENABLING EXTENDED RANGE OF NUMERIC (FLOAT) VALUES
- TimescaleDB 1.7 ( 2.0 )
PostgreSQL, PostgreSQL. — 24 CPU, 64 GB RAM. , , . ~350 15 . .
TimescaleDB , "Faster Method":
- , . , history
- , ,
CREATE TABLE history_new (LIKE history INCLUDING DEFAULTS INCLUDING CONSTRAINTS EXCLUDING INDEXES);
-
SELECT create_hypertable('history_new', 'clock', chunk_interval => 86400);
- history history_new
INSERT INTO history_new SELECT * FROM history;
- history
DROP TABLE IF EXISTS history;
- history_new history
ALTER TABLE IF EXISTS history_new RENAME TO history;
- ( — schema.sql)
CREATE INDEX history_1 in history (itemid,clock);
:
- history
- history_log
- history_str
- history_text
- history_uint
- trends
- trends_uint
, .
github .
finish.sql .
TimescaleDB 5 , 15.
PROFIT.