On 01/23/2018 04:27 PM, Georgios Petasis wrote:
Ok, I have changed MPM back to prefork, and I have changed my apache
configuration to:
<IfModule mpm_prefork_module>
LoadModule rivet_module modules/mod_rivet.so
RivetServerConf MpmBridge prefork
</IfModule>
<IfModule mpm_worker_module>
LoadModule rivet_module modules/mod_rivet.so
RivetServerConf MpmBridge worker
</IfModule>
<IfModule mpm_event_module>
Error "mod_rivet does not support the event mpm."
</IfModule>
And I think I am back to normal. And next time the OS changes the MPM, I
will get at least an error while loading apache...
Best regards,
George
sorry, I forgot that you have established that tdbc and threads don't
get very well together, so you can't safely test your application with
the worker MPM neither, can you?
In a previous message you said that what you observe is a frequent call
to the global_init_script (called by Rivet_VirtualHostsInterps). This
script is evaluated only when threads are started. This happens only
during the child initialization phase. With the prefork bridge the
ChildInit callback itself calls Rivet_VirtualHostsInterps. The worker
bridge calls it from each worker thread, only once in the thread
lifetime. So I think it's strange, if crashes don't occur, that this
code in called so frequently. Do you have clues in the error log?
Unfortunately the documentation is shy of providing implementation
details of the event bridge, but it's stated that as a MPM is based on
the worker MPM having as main difference the way connections are handled.
-- Massimo
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
