Support for Drupal 7 is ending on 5 January 2025—it’s time to migrate to Drupal 10! Learn about the many benefits of Drupal 10 and find migration tools in our resource center.
Updated: Comment #0
Problem/Motivation
Since November 18 or 19, numerous unrelated tests have been failing randomly on the testbots with the following error:
Fatal error: Allowed memory size of 268435456 bytes exhausted (tried to allocate 268435456 bytes) in /var/lib/drupaltestbot/sites/default/files/checkout/core/lib/Drupal/Core/Database/Connection.php on line 336
It is not may always be in Connection.php.
Examples:
https://qa.drupal.org/pifr/test/674688
https://qa.drupal.org/pifr/test/674683
The fatal occurs in completely random tests, sometimes just one fail, sometimes hundreds.
Proposed resolution
Track down and revert the cause of this regression, and then profile to determine how to fix that patch.
Remaining tasks
TBD
API changes
TBD
Comments
Comment #1
xjmComment #2
rfayWe can easily increase the testbot memory_limit, which hopefully could give some breathing space to chase this. Unless it has to do with recursion or something that has nothing to do with actual memory space.
We can make available a dedicated testbot for doing some git bisect action to try to sort this out.
Comment #3
xjmComment #4
xjm@amateescu reverting a suspect commit:
https://drupal.org/comment/8200987#comment-8200987
Comment #5
rfay@amateescu pointed out https://qa.drupal.org/pifr/test/675143 so I grepped on that server and found no sign of the error message (grep -r in /var/log after failing in /var/log/apache2). Then I realized this message must come from run_tests.sh instead of Apache/php
Comment #6
xjmHere's the HEAD failures for the past several days. (Unfortunately these do not include the full log so we can't check the actual PHP error in each one.)
Comment #7
xjmBisecting on the bot:
https://drupal.org/node/2126447#comment-8201283
Comment #8
amateescu CreditAttribution: amateescu commentedContinued: https://drupal.org/node/2126447#comment-8201631
Comment #9
xjmThis is probably the one (from @amateescu). You can see the results of testing it in https://drupal.org/comment/8202607#comment-8202607. The one fail appears to be legitimately a bot issue and not related to the regression, see: https://qa.drupal.org/pifr/test/675568
Comment #10
catchRolled that back. Thanks for the bisecting!