Re: daemontools respin?

Jeremy Hansen <[email protected]> Wed, 9 Jun 2010 10:29:37 -0700
Newsgroups gmane.comp.sysutils.bgware
Message-ID <[email protected]>
On Jun 9, 2010, at 10:20 AM, Bruce Guenter wrote:

> On Wed, Jun 09, 2010 at 09:20:08AM -0700, Jeremy Hansen wrote:
>> This sounds great.  Getting rid of the necessity of /service all together would be great.
>> 
>> This is outside of the realm of supervise itself, but I would love to see a /etc/inittab.d/ type implementation so it becomes unnecessary to alter /etc/inittab directly.  We try to stay as pristine as possible from the vendor distro and patching /etc/inittab somewhat breaks this goal.
> 
> I'm not clear what you're trying to say about /service.  svscan is
> agnostic to what directory the services go into, so putting them into
> /etc/service.d or whatever is no big deal.  I'm not immediately aiming
> at having svscan useable as a replacement for init, but I certainly
> would not object to it either.

I'm talking about the fact that /etc/inittab has to be patched to insert:

SV:2345:respawn:/usr/bin/svscan-start /service
SX:S016:wait:/usr/bin/svscan-stopall /service

It would be cleaner of there was a /etc/inittab.d/ that would include something like /etc/inittab.d/supervise containing the above lines.

This is more for the maintainers of the initscripts package, which is why I said it didn't apply to supervise itself.

Although what you're saying about using a /etc/service.d/ style location for supervise scripts sounds like a good idea as well.  I realize the location of /service isn't static.  This is where I disagree with DJB in terms of just randomly creating new top level directories.  Supervise services really should go into /etc like SysVinit.

I'm not *as* concerned with things properly managed by RPM, but it would still be nice to try to conform to the standards.  I like clean installs.

Thanks
-jeremy

> 
> -- 
> Bruce Guenter <[email protected]>                http://untroubled.org/