Lennart Poettering [2015-02-03 17:29 +0100]: > Hmm, why precisely does this stall for 90s?
The current transaction has final.target and all other jobs which need to be shut down. One of these now trigger "systemctl reload postfix.service", but that reload isn't going to actually run in the same transaction but in the next one. OTOH systemctl reload waits for the reloading to finish, thus we have a deadlock. > Isn't this a case where people should just use "--no-block"? Kind of. Not using this is the right thing while the machine is running, so that the reload is actually done after calling systemctl reload, and you can go on using postfix or whatever. --no-block should help during shutdown, or early boot (same principal bug, but slightly different patch, see http://bugs.debian.org/624599). So every time you call reload you'd have to check whether or not you are in early boot/during shutdown, or in the running system, and conditionally use --no-block. However, as such scripts should never call systemctl directly, but "service foo reload" (to work with other init systems or chroot), it would be also possible to move that check there, and conditionally add --no-block. It would just be another thing that every distro has to re-discover :-) Thanks, Martin -- Martin Pitt | http://www.piware.de Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org) _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
