You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@trafficserver.apache.org by GitBox <gi...@apache.org> on 2021/02/24 22:27:54 UTC

[GitHub] [trafficserver] ywkaras edited a comment on issue #7546: LogObject::_checkout_write using a lot of CPU on many core servers

ywkaras edited a comment on issue #7546:
URL: https://github.com/apache/trafficserver/issues/7546#issuecomment-785425829


   One thing I'm seeing is that, when one thread is blocked waiting on this memory allocation:  https://github.com/apache/trafficserver/blob/564663b9b6888c8334ec8077b5583da9b192c3c2/proxy/logging/LogObject.cc#L399 , any other  thread trying to reserve (checkout) buffer space from the same LogObject will keep getting the return value LB_RETRY from LogBuffer::writer_checkout(), and will thus busy wait.  So we could be indirectly seeing the memory allocation blocking problem.  If this is the case, it would likely help to execute a thread yield when the LB_RETRY return value is returned.
   
   I think I need to reproduce the problem with a baseline build, then see if this change improves things.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org