On 07/17/2011 08:26 PM, Regid Ichira wrote: > Package: cron > Version: 3.0pl1-118 > Severity: normal > Tags: patch > > init.d/cron used to be one of the last scripts to run. With > parallel boot sequence, this is not necessarily true. In order > to preserve the semantics of the @reboot time specification, I > suggest to:
This would indeed permanently fix a number of issues we encountered with regards to bootup, eg the recent bugs concerning network and DNS availability, but also the availability of name caching daemons and who-knows-what-else. However, I'm absolutely certain of having considered this solution in the past and ultimately dismissing it -- I currently can't recall why, though. I'll leave this bug open until I remember why... > --- cron 2011-07-17 21:04:04.000000000 +0300 > +++ cron.new 2011-07-17 21:04:34.000000000 +0300 > @@ -3,7 +3,7 @@ > # > ### BEGIN INIT INFO > # Provides: cron > -# Required-Start: $remote_fs $syslog $time > +# Required-Start: $remote_fs $syslog $time $all > # Required-Stop: $remote_fs $syslog $time > # Should-Start: $network $named slapd autofs ypbind nscd nslcd > # Should-Stop: $network $named slapd autofs ypbind nscd nslcd > > As an aside, I do not understand why one should list > $remote_fs $syslog $time > when $all is stated. I mean, doesn't $all includes those services? > Having $all listed with other services is copied from init.d/rc.local. > Can it be that $all is insserv's specific, and there was an attempt > to allow other boot scripts' enumerating facilities? I'm not sure I entirely follow, but I read insserv(8) as Should-Start: $all being correct (ie, without $remote_fs $syslog $time). Christian
signature.asc
Description: OpenPGP digital signature
