Bug #83451

DatabaseRowsUpdateWizard Out of Memory on big tables

Added by Alexander Opitz over 1 year ago. Updated about 2 months ago.

Status:
Needs Feedback
Priority:
Should have
Assignee:
-
Category:
Install Tool
Start date:
2018-01-02
Due date:
% Done:

0%

TYPO3 Version:
8
PHP Version:
Tags:
Complexity:
Is Regression:
Sprint Focus:

Description

The "database migrations on single rows" wizard selects complete tables, which leads to memory exhausted on large tables (on my system 1,6 million rows) as all rows get selected. Normally the rows should only be read from database while one row is processed after the other one (as we don't use something like fetchAll) but there are PHP issues around the buffering on MySQLi and pdo_mysql drivers.
See https://secure.php.net/manual/en/mysqlinfo.concepts.buffering.php about this (And special one in MySQLi see https://github.com/doctrine/dbal/blob/master/lib/Doctrine/DBAL/Driver/Mysqli/MysqliStatement.php#L192).

So we should move from select all to select only 1(0).000 rows at once. This will be slower but helps in this OoM issue.


Related issues

Related to TYPO3 Core - Task #86200: Migrate RowUpdater wizards Closed 2018-09-08

History

#1 Updated by Alexander Opitz over 1 year ago

Shortest complete answer: https://stackoverflow.com/questions/13728106/unexpectedly-hitting-php-memory-limit-with-a-single-pdo-query

But we can't switch from buffered to unbuffered read, as we do queries while iterating over the rows.

#2 Updated by Christian Kuhn about 1 year ago

  • Status changed from New to Needs Feedback

unsure on how to proceed. any ideas? note: in v9, the row updater can be called via cli, this may relax the issue a bit.

#3 Updated by Christian Kuhn about 1 year ago

  • Related to Task #86200: Migrate RowUpdater wizards added

#4 Updated by Kevin Ditscheid about 2 months ago

Just stumbled into this while updating a huge instance.
If I set $connectionForTable->getWrappedConnection()->setAttribute(\PDO::MYSQL_ATTR_USE_BUFFERED_QUERY, false); with driver pdo_mysql, I can step over the statement execution, but the open unbuffered connection does not allow for another query.
```
An exception occurred while executing 'SELECT COUNT FROM `sys_registry` WHERE (`entry_namespace` = ?) AND (`entry_key` = ?)' with params ["installUpdateRows", "rowUpdatePosition"]: SQLSTATE[HY000]: General error: 2014 Cannot execute queries while other unbuffered queries are active. Consider using PDOStatement::fetchAll(). Alternatively, if your code is only ever going to run against mysql, you may enable query buffering by setting the PDO::MYSQL_ATTR_USE_BUFFERED_QUERY attribute.
```
I am unsure how to solve this. The table sys_file_metadata has more than 1.5 Million entries alone.
I mean, an Idea to solve this could be, that we fetch the whole row inside of the for-loop and only fetch uids beforehand.
This would however cause a massive increase in database connections, 1.5 Million for the sys_file_metadata alone in my case.

Also available in: Atom PDF