Weird behaviour after doctrine-bundle update 2.0.6

[acasademont]

Just to keep track of it, seems like the new doctrine-bundle default behaviour interferes with our clear or reset strategy we have in the Symfony bootstrap.

https://github.com/doctrine/DoctrineBundle/issues/1114

[dnna]

That looks strange indeed. I'll try to have a look in the next few days, as this will probably affect my projects as well after upgrading. The PHP-PM code itself doesn't look wrong, it only resets if the entity manager is not open (e.g. an exception occurred that closed it), otherwise it only clears the existing one.

[dnna]

Btw, if DoctrineBundle now performs its own reset we should probably remove our reset logic in PHP-PM's postHandle or make it conditional on Doctrine\Bundle\DoctrineBundle\Registry not having ResetInterface. Otherwise we are resetting twice now.

[acasademont]

yes, exactly, it's as easy as to check if the doctrine service is implementing the ResetInterface or not

[mathieudz]

I'm having memory leaks after upgrading from doctrine-bundle 2.0.2 to 2.0.6. Workers reach max memory after 500 requests while the workers are only restarted after 1000 requests. I've downgraded it back to 2.0.2

[acasademont]

Interesting side effect, I didn’t even notice. I’m going to check our production logs. In any case, we too reverted to 2.0.2

On Tue, 24 Dec 2019 at 22:00, Mathieu De Zutter [email protected] wrote:

I'm having memory leaks after upgrading from doctrine-bundle 2.0.2 to 2.0.6. Workers reach max memory after 500 requests while the workers are only restarted after 1000 requests. I've downgraded it back to 2.0.2

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/php-pm/php-pm-httpkernel/issues/161?email_source=notifications&email_token=AAGJNPO2VVEMIX5LNGELZVTQ2JZ6PA5CNFSM4J6VOTZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEHTTT5I#issuecomment-568801781, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAGJNPMV5SVFXISPQDG744TQ2JZ6PANCNFSM4J6VOTZQ .

[acasademont]

@mathieudz didn't see any memory limit crashes on our logs. Still, we should also keep an eye on it.