Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 09 May 2015 20:24:26 +0100
From:      Jonathan de Boyne Pollard <J.deBoynePollard-newsgroups@NTLWorld.com>
To:        FreeBSD Hackers <freebsd-hackers@freebsd.org>
Subject:   Re: nosh version 1.14
Message-ID:  <554E5EEA.7020901@NTLWorld.com>
In-Reply-To: <554E53EF.4080600@NTLWorld.com>
References:  <54430B41.3010301@NTLWorld.com> <54B86FD5.3090203@NTLWorld.com> <554E53EF.4080600@NTLWorld.com>

next in thread | previous in thread | raw e-mail | index | archive | help
nosh is now up to version 1.14

* http://homepage.ntlworld.com./jonathan.deboynepollard/Softwares/nosh.html

That was the source package changelog.  Now the discussion.

I've been running FreeBSD 10 as an entirely nosh-managed system for 
quite a while, now.  But I still had some ad hoc mechanisms for starting 
the user-space virtual terminals by logging in on a kernel virtual 
terminal and running some scripts.  I'm now running user-space virtual 
terminals as proper services, under the service manager, so they come up 
at bootstrap unattended.  As you can see from the changelog, I've fixed 
a font bug in the terminal emulator and eliminated some of the very few 
remaining to-do items for it.

Input-activated log services, as started by svscan (a.k.a. 
service-dt-scanner), are now gone.  There's a minor story behind this.  
Monitoring real systems with nagios militated against it. The nagios 
check would report many of the log services as down, because of course 
the logged services hadn't done anything worth logging and hadn't 
activated the loggers.  For my use, good nagios checks were more 
important than the input activation, which was only ever an experiment 
in any case.  So now svscan no longer uses it. On that score, there is 
also a small improvement in the nagios checking tool, which allows one 
to specify the minimum time that a service has to be in the running 
state before it will be considered OK, nagios-wise.

Similarly, there's now a command to tell service-manager to unload a 
service when it next enters the stopped state.  The service manager will 
disconnect itself from the control FIFOs and the supervise/ directory, 
and forget all about the service.  This allows a service bundle to be 
cleanly removed from the system in its entirety.

I've added a command to convert /etc/fstab into mount and fsck 
services.  More is planned here, including integration with the rc.conf 
and ttys conversion utilities under an umbrella of some sort, details 
yet to be decided.

The packaging changes are a big deal in the Linux world, but they aren't 
really relevant in the BSD world.  This just leaves a whole load of new 
service bundles, which are a message all to themselves.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?554E5EEA.7020901>