On Tue, Nov 13, 2012 at 8:59 PM, Peter Danenberg <p...@roxygen.org> wrote:
> mod_fcgid doesn't appear to have an external process manager; compare > this process tree from mod_fastcgi where the access checker > (872695a60a6d82aa9fc041bae44803af) is a child of fcgi is a child of > httpd: > > /usr/local/apache2/bin/httpd -k start > \_ /usr/local/apache2/bin/fcgi- -k start > | \_ 872695a60a6d82aa9fc041bae44803af > > to this one from mod_fcgid where the access checker is a direct > descendant of httpd: > > /usr/local/apache2/bin/httpd -k start > | \_ 872695a60a6d82aa9fc041bae44803af > mod_fcgid has a daemon-child of the httpd parent that spawns FastCGI applications. The backtrace will be something extremely similar to apr_wait_for_io_or_timeout() < procmgr_peek_cmd() <- pm_main() <- create_process_manager() <- procmgr_post_config() <- fcgid_init() <- ap_run_post_config() <- main() > I've attached straces from mod_fastcgi's process manager; as well as > straces from 872695a60a6d82aa9fc041bae44803af running under mod_fcgid > and mod_fastcgi, if that's helpful. > > Do you still have the original httpd-strace for mod_fcgid? If so, grep > for access.scm; you'll see the script there. > Did you have a chance to compare the activity of the FastCGI app in both circumstances, to see how the interaction differed? > > PS. I've since compiled mod_fastcgi against httpd-2.4.3; the only > reason I was screwing around with mod_fcgid was that mod_fastcgi had > stopped compiling against 2.4.3. > > It may be time to consider mod_fcgid abandonware, since I haven't > heard from the authors; and go back to using mod_fastcgi in > production. > Which version of mod_fcgid are you using? Modern mod_fcgid (2.3.x) is supported on this mailing list and in ASF Bugzilla. > > On Sun, Nov 11, 2012 at 8:41 PM, Peter Danenberg <p...@roxygen.org> > wrote: > > > > > Thanks for looking into this, Jeff; I've attached an strace with > > > mod_fcgid (strace-fcgid), one with mod_fastcgi (strace-fastcgi) and > > > one with mod_fcgid and relative time stamps so you can see where it > > > hangs (strace-fcgid-with-relative-time). > > > > > > > Ouch. Is your script running in there somewhere? I just saw httpd > > activity but no activity of your script. (s/script/application/) > > > > Perhaps the most straightforward way to get a trace is to start httpd, > > identify the mod_fastcgi or mod_fcgid daemon process, start strace -fp > > DAEMON-PID, and then make the request. > > > > How to identify the mod_fcgid daemon: At startup, it logs a message at > > LogLevel info which says "mod_fcgid: Process manager XXXX started", where > > XXXX is the pid. > > > > mod_fastcgi? I don't know. But you may be very used to gdb and strace > > before this is over, so maybe you can pick the child that doesn't look > like > > the others ;) > > > > --/-- > > > > OTOH, how much effort is required to get your chicken-scheme script > running > > on Linux? > > > > > > > > > > > > It looks like the immediate call is: > > > > > > rt_sigtimedwait(~[ILL TRAP ABRT BUS FPE KILL SEGV USR2 PIPE CONT > > > STOP SYS RTMIN RT_1], NULL, NULL, 8) = 1 > > > > > > > > > > > > > > > > On Sunday, November 11, 2012, Peter Danenberg wrote: > > > > > > > > > Digging a little deeper, mod_fcgid hangs on line 318 of > > > > > fcgid_bridge.c: > > > > > > > > > > ap_scan_script_header_err_core(r, sbuf, getsfunc_fcgid_BRIGADE, > > > > > brigade_stdout) > > > > > > > > > > I don't know enough about ap_scan_script_header_err_core yet to > figure > > > > > out why this might be the case. > > > > > > > > > > > > Can You compare an strace/truss/dtruss of your script running with > the > > > > alternative modules? > > > > > > > > > > > > > > > The following trivial script,[1] which simply returns a 200 > header, > > > > > > hangs under FcgidAccessChecker: > > > > > > > > > > > > (call-with-dynamic-fastcgi-query > > > > > > (λ (query) > > > > > > (display-status-&c.))) > > > > > > > > > > > > The same script works fine under mod_fastcgi with > > > > > > FastCgiAccessChecker; or handled as an fcgid-script or > > > fastcgi-script. > > > > > > > > > > > > The output from error_log is as follows: > > > > > > > > > > > > [Sun Nov 11 01:39:44 2012] [info] mod_fcgid: server > > > > > > localhost:/usr/local/var/html/scm/api/access.scm(15965) > > > > > > started > > > > > > > > > > > > [Sun Nov 11 01:40:24 2012] [warn] [client 127.0.0.1] mod_fcgid: > > > read > > > > > > data timeout in 40 seconds > > > > > > > > > > > > [Sun Nov 11 01:40:24 2012] [error] [client 127.0.0.1] Premature > > > end of > > > > > > script headers: access.scm > > > > > > > > > > > > [Sun Nov 11 01:40:24 2012] [warn] [client 127.0.0.1] mod_fcgid: > > > user > > > > > > (null) access check failed, respond 500, URI > /scm/api/access.scm > > > > > > > > > > > > [Sun Nov 11 01:40:26 2012] [debug] fcgid_pm_main.c(357): > mod_fcgid: > > > > > > gracefully terminated 1 processes > > > > > > > > > > > > How can I diagnose the problem? I find it interesting that it > works > > > > > > under mod_fastcgi but not mod_fcgid; I'm not sure whether that > > > implies > > > > > > the problem is in mod_fcgid, however. > > > > > > > > > > > > > > > > > > Footnotes: > > > > > > [1] The script is written in Chicken Scheme, by the way. > > > > > > > > > > > > > > > > > > > --------------------------------------------------------------------- > > > > > > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > > > <javascript:;> > > > > > > For additional commands, e-mail: users-h...@httpd.apache.org > > > <javascript:;> > > > > > > > > > > > > > > > > > > > > > > --------------------------------------------------------------------- > > > > > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > <javascript:;> > > > > > For additional commands, e-mail: users-h...@httpd.apache.org > > > <javascript:;> > > > > > > > > > > > > > > > > > > -- > > > > Born in Roswell... married an alien... > > > > http://emptyhammock.com/ > > > > > > > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > > > For additional commands, e-mail: users-h...@httpd.apache.org > > > > > > > > > > > -- > > Born in Roswell... married an alien... > > http://emptyhammock.com/ > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > For additional commands, e-mail: users-h...@httpd.apache.org > -- Born in Roswell... married an alien... http://emptyhammock.com/
