Hi, modify /etc/nscd.conf that way, that it contains an error. E.g. add an unsupported option. Now (re)start nscd and check status: # /etc/init.d/nscd start && /etc/init.d/nscd status * Starting Name Service Cache Daemon ... [ ok ] * status: crashed If we would add "--wait N" to the start-stop-daemon call, we would notice the error. But please see comment 0 in bug 387107 regarding a good "N" value. Reproducible: Always
I am unsure if we should set the bug to depend on bug 387107. It would be nice to have a "dynamic" check, but I think a value of 500-1000ms (0.5-1s) should be fine for most setups. If you are using nscd in a large LDAP environment with persistent data you may want to increase the time. Therefore we could make it configurable, e.g: # cat /etc/conf.d/nscd NSCD_CHECKAFTER_N_MS=1000 # cat /etc/init.d/nscd NSCD_CHECKAFTER_N_MS=${NSCD_CHECKAFTER_N_MS:-1000} ... start-stop-daemon --start --quiet --background --wait ${NSCD_CHECKAFTER_N_MS} \ --exec /usr/sbin/nscd --pidfile ${pidfile} \ -- --foreground $secure I think you get the idea :)
Created attachment 364296 [details] New nscd runscript Hi, I am attaching a new runscript for nscd. Changes: 1) Added "syslog" to USE, because there are situation when nscd should write to syslog (see db_log() in dbg_log.c). 2) Changes "/var/run" paths to "/run" (no important change because "/var/run" is a symlink to "/run", but "/run" seems to be the new default) 3) checkconfig() will now be called from start_pre() 4) start() and stop() isn't needed, we can use the default functions 4.1) Changes from 4 removes "--secure" support, too. This shouldn't be a problem, because "-S" (--secure) was removed from nscd: # nscd -S passwd nscd: secure services not implemented anymore I did not verify when it was removed, so this runscript is only tested with sys-libs/glibc-2.17 and later. 4.2) Without the "--background" call we don't need "--wait" (=this will solve the problem described in this bug report). Because we are now using the default start and stop functions, we now also support the "SSD_STARTWAIT" environment variable. So in situation described in comment 1, one could set SSD_STARTWAIT in "/etc/conf.d/nscd" 4.3) Not sure why we set "--foreground". According to http://thread.gmane.org/gmane.comp.lib.glibc.alpha/17144 this feature was added for systemd. From my tests it seems like we don't need it. 5) The declaration of "pidfile" allows the built-in status function proper checks. 6) I added extra functions "flush_all", "flush_hosts", "flush_group", "flush_passwd", "flush_netgroup" and "flush_services" which will flush the named caches. 7) I added descriptions.
Created attachment 364338 [details] New nscd runscript I updated the attachment (changed "syslog" to "logger").
Updated script with minor cosmetic changes added to the glibc-2.26 patchset (will be in patchset 1 and later)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=02056778ea5961e77a59a7a246b355c1225c7404 commit 02056778ea5961e77a59a7a246b355c1225c7404 Author: Andreas K. Hüttel <dilfridge@gentoo.org> AuthorDate: 2017-11-12 12:28:38 +0000 Commit: Andreas K. Hüttel <dilfridge@gentoo.org> CommitDate: 2017-11-12 14:15:28 +0000 sys-libs/glibc: Re-add keywords to glibc 2.26 Bug: https://bugs.gentoo.org/492814 Bug: https://bugs.gentoo.org/622694 Bug: https://bugs.gentoo.org/617938 Bug: https://bugs.gentoo.org/466176 Bug: https://bugs.gentoo.org/628768 Bug: https://bugs.gentoo.org/637016 Bug: https://bugs.gentoo.org/636934 Bug: https://bugs.gentoo.org/381391 Bug: https://bugs.gentoo.org/636158 Package-Manager: Portage-2.3.13, Repoman-2.3.4 sys-libs/glibc/glibc-2.26-r3.ebuild | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)}
Added to the 2.25 patchset (patchlevel 13 and later)
Also fixed in 2.25-r10
(In reply to Andreas K. Hüttel from comment #7) > Also fixed in 2.25-r10 Is now stable everywhere
