Latest php was released with new init script for php-fpm. Which propably wasnt *even* tested because it failing on start and hard to overlook it... Script waiting for pidfile however pidfile will be never created because there is no -m (--makepid) switch to start-stop-deamon in script. Also even with -m pidfile contain wrong pid, php-fpm master process is created after it with highter pid - In that case 'restart' will fail to stop php-fpm (because pid does not exist) and will start another php-fpm. Anoter careless upgrade to php. Reproducible: Always Steps to Reproduce:
(In reply to comment #0) > Latest php was released with new init script for php-fpm. Which propably wasnt > *even* tested because it failing on start and hard to overlook it... > > Script waiting for pidfile however pidfile will be never created because there > is no -m (--makepid) switch to start-stop-deamon in script. Also even with -m > pidfile contain wrong pid, php-fpm master process is created after it with > highter pid - In that case 'restart' will fail to stop php-fpm (because pid > does not exist) and will start another php-fpm. > Please try to merge your fpm config with the one shipped with 5.3.4 (which should be in the tree now re bug 349179). let me know how that works.
(In reply to comment #0) > Script waiting for pidfile however pidfile will be never created because there > is no -m (--makepid) switch to start-stop-deamon in script. Also even with -m > pidfile contain wrong pid, php-fpm master process is created after it with > highter pid - In that case 'restart' will fail to stop php-fpm (because pid > does not exist) and will start another php-fpm. Another problem comes from the wrong usage of the PHPSLOT variable. You can fix it by changing PHPSLOT declaration as follows: PHPSLOT=$(basename $(eselect php show fpm)) Yes, a careless upgrade indeed.
(In reply to comment #2) > Another problem comes from the wrong usage of the PHPSLOT variable. > You can fix it by changing PHPSLOT declaration as follows: > PHPSLOT=$(basename $(eselect php show fpm)) How's that? eselect php show fpm gives "php5.3" for me. What does it output for you?
Ah, so now php-fpm init depends on pid variable from php-fpm.conf without any ewarn/einfo/notice about it. Such changes *should be* docummented in ebuild's postinst or something... http://bugs.gentoo.org/show_bug.cgi?id=343671#c8 Config is something very individual, uncommeting pid variable is something easy to overlook while dispatch-conf.
Uncommenting "pid" global option in php-fpm.conf works for me. Thanks for info.