Home/ Questions/Q 7890161
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-03T06:16:58+00:00

I have some kind of (basic) logging with user actions in a postgreSQL database.

  • 0

I have some kind of (basic) logging with user actions in a postgreSQL database.
In order to gain performance, I execute all log inserts asynchronously to let the script continue and not wait until log entry is created.

I use prepared statements everywhere to prevent SQL injections and load them in an as-needed basis.

The problem comes when there are pending results to be fetched from a previous async query when I prepare a statement. (PostgreSQL says there are pending results to be fetched prior to prepare a new statement)

So as a workarround, I gather all pending results (if any) and ignore them to make PHP and PostgreSQL happy before preparing any statement.

But with that workarround (as I see it), I miss the performance I could gain by executing asyncronously as I have to gather the results anyway.

Is there any way to asynchronously execute a prepared statement and deliberatelly tell postgres to ignore results?

Inside my PostgreSQL class, I am calling prepared statements with

pg_send_execute($this->resource, $name, $params);

and prepairing them with 

//Just in case there are pending results (workarround)
while (pg_get_result($this->resource)!==FALSE);
$stmt = pg_prepare($this->resource, $stmtname, $query);

Any help will be apreciated.

UPDATE: All asynchronous queries I am using are only INSERT ones, so it should be safe (theoretically) to ignore their results.

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    The only thing that is asynchronous is your communication with PostgreSQL server – your database has to process everything sequentially.

    My proposal:

    If you have to use PostgreSQL for logging, use a separate database connection for logging purposes and get a connection pool sitting between your script and database – auth in PostgreSQL is costly and takes some time, this will cut it down. Acquiring a second connection will take some time, but if you use this method it will be faster than one without connection pool.

    Depending on your reliability requirements you should use autocommit (to never lose a log entry when PHP crashes). You may want to use an UNLOGGED table (available since PostgreSQL 9.1) if you don’t care about reliability on databse end (faster inserts as your data skips WAL) or if you don’t use replication or don’t need to have logs replicated.

    As a speed optimization, your log table should have no indexes because they would have to be updated on each insert. If you need them, create a second table and move data in a batch (every X minutes or every hour).

    • 0