This is a document for PGroonga 1.X. See PGroonga 2.x document when you're using recent PGroonga.

pgroonga.flush function


pgroonga.flush function ensuring writing changes only in memory into disk. Normally, you don't need to this function because it's done automatically. But you may need to use this function when you want to prevent breaking PGroonga indexes on crash or force shutdown.

Normally, users shouldn't shut down server forcibly but some users do in some cases. For example, Windows update may restart Windows server unexpectedly.

If PostgreSQL with PGroonga is shut down forcibly, changes only in memory may be lost. If you call pgroonga.flush function before force shutdown, there are no changes only in memory. It means that PGroonga indexes aren't broken even if users shut down PostgreSQL with PGroonga forcibly.

If there are many changes only in memory, pgroonga.flush may take a long time. It's depend on write performance of your disk.


Here is the syntax of this function:

bool pgroonga.flush(pgroonga_index_name)

pgroonga_index_name is a text type value. It's an index name to be flushed. The index should be created with USING pgroonga.

pgroonga.flush returns always true. Because if pgroonga.flush is failed, it raises an error instead of returning result.


Here are sample schema and data. In the schema, both search target data and output data are index target columns:

  id integer,
  title text,
  content text

CREATE INDEX pgroonga_terms_index
          ON terms
       USING pgroonga (title, content);

You can flush pgroonga_terms_index related changes only in memory by the following pgroonga.flush call:

SELECT pgroonga.flush('pgroonga_terms_index');
--  flush 
-- -------
--  t
-- (1 row)

If you specify nonexistent index name, pgroonga.flush raises an error:

SELECT pgroonga.flush('nonexistent');
-- ERROR:  relation "nonexistent" does not exist