Mailing List Archive: Apache: Bugs

[Bug 48110] Doc of mod_fcgid: "PHP child process management" appears incorrect

Index | Next | Previous | View Threaded

bugzilla at apache

Nov 3, 2009, 6:58 AM

Post #1 of 3 (157 views)
Permalink

[Bug 48110] Doc of mod_fcgid: "PHP child process management" appears incorrect

https://issues.apache.org/bugzilla/show_bug.cgi?id=48110

--- Comment #1 from Jeff Trawick <trawick[at]apache.org> 2009-11-03 06:58:53 UTC ---
Hi there,

I'm glad to see that someone is reading the new docs, and I must confess that
the PHP-related information is particularly deserving of scrutiny due to the
lack of PHP documentation in this area as well as the wide distribution of
conflicting information about behavior with mod_fcgid.

The heart of the matter is that mod_fcgid will not route a request to a process
it has not created, and will not route to a process if a request previously
routed to it is still in progress. Thus, if the PHP process created by
mod_fcgid is idle or has an idle child process because it has passed a current
request to a child, mod_fcgid still will not send another request to it.

I think this incomplete+incorrect information

"... which will only route requests to application processes it has spawned.
Any child processes created by PHP will be unused ..."

can be replaced with

"... which will only route requests to application processes it has spawned,
and is not aware that PHP may have passed active requests to another PHP
process. Any child processes created by PHP will not be effectively utilized
..."

Does this make sense?

Do you have empirical evidence to the contrary?

Thanks!

--
Configure bugmail: https://issues.apache.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe[at]httpd.apache.org
For additional commands, e-mail: bugs-help[at]httpd.apache.org

bugzilla at apache

Nov 3, 2009, 7:49 AM

Post #2 of 3 (145 views)
Permalink

[Bug 48110] Doc of mod_fcgid: "PHP child process management" appears incorrect [In reply to]

https://issues.apache.org/bugzilla/show_bug.cgi?id=48110

--- Comment #2 from HWS <strube[at]physik3.gwdg.de> 2009-11-03 07:49:28 UTC ---
Well, I think that the PHP child processes are simply irrelevant to mod_fcgid;
the parent PHP is the only one seen and treated like a single PHP process. I
do not understand "if the PHP process created by mod_fcgid is idle or has an
idle child process because it has passed a current request to a child,
mod_fcgid still will not send another request to it." How can the parent PHP
process possibly be considered idle by mod_fcgid as long as one of its children
is still processing a request?
But you really make an important point: "will not route to a process if a
request previously routed to it is still in progress." This could mean that a
PHP parent with several children would not be faster than a single PHP process.
(I have not really measured this, also don't know whether mod_fastcgi and
mod_fcgid act differently in this respect.) Then your replacement would make
sense.
However, in my scenario with APC extension, the child management by a PHP
parent is even then advantageous, since crashed children are replaced without
losing the content of the APC shared-memory cache (which is created during
parent startup and is shared with the children).

--
Configure bugmail: https://issues.apache.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe[at]httpd.apache.org
For additional commands, e-mail: bugs-help[at]httpd.apache.org

bugzilla at apache

Nov 4, 2009, 5:26 AM

Post #3 of 3 (137 views)
Permalink

[Bug 48110] Doc of mod_fcgid: "PHP child process management" appears incorrect [In reply to]

https://issues.apache.org/bugzilla/show_bug.cgi?id=48110

--- Comment #3 from HWS <strube[at]physik3.gwdg.de> 2009-11-04 05:26:37 UTC ---
More info re "This could mean that a PHP parent with several children would not
be faster than a single PHP process.(I have not really measured this, also
don't know whether mod_fastcgi and mod_fcgid act differently in this respect.)"

I just compared mod_fastcgi and mod_fcgid using PHP_FCGI_CHILDREN = 1 or 2 in
both cases and benchmarking with "ab -n 50 -c 2 http://...." on a
multiprocessor SPARC server (Solaris 10). Indeed, with mod_fastcgi, the time
was halved for PHP_FCGI_CHILDREN=2, whereas this was not the case with
mod_fcgid! Thus, mod_fastcgi is superior in this respect. (On a
single-processor machine, this will probably be irrelevant.) Configuration was:
FcgidIPCDir /some/dir
FcgidProcessTableFile /some/file
FcgidIdleTimeout 0
FcgidMaxProcesses 1
FcgidMaxProcessesPerClass 1
FcgidMinProcessesPerClass 1
FcgidInitialEnv PHPRC /some/path
FcgidInitialEnv PHP_FCGI_CHILDREN 2 # or 1 or 0
FcgidInitialEnv TZ MET
FcgidFixPathinfo 1

--
Configure bugmail: https://issues.apache.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe[at]httpd.apache.org
For additional commands, e-mail: bugs-help[at]httpd.apache.org

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact lists@gossamer-threads.com