Not logged in
Log in now
Create an account
Subscribe to LWN
LWN.net Weekly Edition for May 16, 2013
A look at the PyPy 2.0 release
PostgreSQL 9.3 beta: Federated databases and more
LWN.net Weekly Edition for May 9, 2013
(Nearly) full tickless operation in 3.10
Shuttleworth: Quality has a new name
Posted Apr 24, 2012 18:33 UTC (Tue) by Cyberax (✭ supporter ✭, #52523)
Gentoo doesn't really support upstart - it can be used in SysV compatibility mode, but it's not being actively advanced.
Posted Apr 24, 2012 18:55 UTC (Tue) by jspaleta (subscriber, #50639)
Seriously I have no idea what the state of the art is. Is Ubuntu using upstart native job definitions for services like sshd or apache right now?
Posted Apr 24, 2012 19:56 UTC (Tue) by drag (subscriber, #31333)
Pretty much all the default services have been moved over to using it natively and a significant portion of packaged services have also. It has made a significant impact on the character of the OS.
Posted Apr 24, 2012 20:04 UTC (Tue) by tpo (subscriber, #25713)
$ ls /etc/init|wc -l
$ lsb_release -r
$ ls /etc/init|grep apache
$ ls /etc/init|grep ssh
Posted Apr 24, 2012 20:04 UTC (Tue) by misc (subscriber, #73730)
but there is for ssh :
Upstart jobs are stored in /etc/init/. There is lots of them on a lucid lynx system :
$ ls /etc/init | wc -l
But that's mostly replacement for the old /etc/rcS.d/ system.
I guess there is more of them on a up to date ubuntu ( for example, I found cobbler in precise, and I guess there is ).
In fact, the real issues is not "using upstart vs using systemd".
The real issue is :
- switching to systemd
- keeping upstart
Switching to systemd is not just a upgrade when coming from upstart, since you cannot just replace upstart jobs by systemd jobs. So I can see why the Ubuntu team thought it would not be worth. They seems to have suffered from switching ( http://undacuvabrutha.wordpress.com/2011/04/29/why-ubuntu... ). So I can see why they prefer to focus their resources to more differentiating features than a init system.
That's transparent to users, and while I definitely prefer systemd ( cause this is much easier for packagers and sysadmins, and much cleaner with the template system for stuff like openvpn ), if desktop users are not gonna see much, this is maybe not the right time to do a big change again.
I do not think this will be a huge burden on Canonical since Google also use it and they have enough people to take care of that.
Posted Apr 24, 2012 21:33 UTC (Tue) by HelloWorld (guest, #56129)
Posted Apr 24, 2012 22:42 UTC (Tue) by rahvin (subscriber, #16953)
If they are counting on Google pulling the weight they are making a mistake (unless they have intimate knowledge of Google's plans that is). Because IMO trusting Google to do anything is asking for trouble. They've got to be the most bi-polar company in existence. The only part of their business that's consistently the same is the advertising arm (and we don't see the internal on that so it could be equally troublesome as the rest). Personally I'm not sure how I would bet on a wager of Google throwing out all previous work and starting over in some radically different direction.
Posted Apr 26, 2012 0:16 UTC (Thu) by mrons (subscriber, #1751)
What about RedHat 6?
In a strange twist, Redhat will need to support upstart for several years now.
Posted Apr 26, 2012 0:27 UTC (Thu) by dlang (✭ supporter ✭, #313)
this may impact the 'obvious' move to systemd in RHEL7, it depends on how disruptive this is to sysadmins and if RedHat is willing to change init packages twice in two releases.
I think that people assume that if something is in Fedora, it will be in the next RHEL. Even if Fedora is the testing ground for RHEL (something disputed by Fedora people regularly), there's still time for the test to fail (for some definition of fail)
Posted Apr 26, 2012 1:11 UTC (Thu) by jspaleta (subscriber, #50639)
Upstart was introduced in Fedora (just as it was in Ubuntu) running in a sysVinit compatibility mode...which created upstart native jobs that mimicked traditional runlevels to run everything else.
In Fedora i'm pretty sure none of the installable services...such as sshd ever had upstart "native" job files and were run as traditional initscripts under the compatibility mode. I expect RHEL 6 to be using upstart in the same way. Which is to say...as close to a replacement of sysvinit and no further than that.
And since I asked the question originally I've looked over the open upstart bugs that have bubbled up in the context of other discussions and it seems upstart has some long standing issues that prevent conversion of a number of significant services like apache and other forking services.
Bug from 2010:
This is pretty much a showstopper for "native" upstart process control across an important swath of daemon (especially for the cloud..cough cough.) I am NOT saying this is unfixable. The original lead developer clearly thought it was fixable...in 2010..by moving away from ptrace in how upstart did its magic. But its NOT FIXED going on a year and a half later to my understanding.
This really brings home the point I think that Canonical has a problem keeping up with their own project development long term. Canonical's choice to keep upstart would make more sense if they were actively cultivating upstart so it could be a full init replacement so all services could gain access to the enhanced event based job control. But they aren't. They made a push to "go native" for as much as was feasible..but they haven't fixed upstart to be able to take it further and complete the work. That's a problem. The longevity of that bug speaks to focus and to quality more than perhaps a lot of people inside the Canonical fenceline realize.
An init replacement system that can't be used to transition from the legacy sysvinit scripts to get access to advanced job flow control features sort of defeats the point a little bit. Especially for a company that is really pushing the server motto. But hey if they want to stick their heels in the sand they are only hurting their own long term prospects. It's always a bit sad to watch it happen. Hopefully they'll double down and put some more engineering manpower into getting upstart polished off instead of letting it linger a slow death.
Posted Apr 26, 2012 11:16 UTC (Thu) by ovitters (subscriber, #27950)
Posted Apr 28, 2012 23:23 UTC (Sat) by speedster1 (subscriber, #8143)
Posted Apr 28, 2012 23:42 UTC (Sat) by Cyberax (✭ supporter ✭, #52523)
Though RHEL7 should be fairly soon - in 2014 or so :)
Posted May 1, 2012 17:05 UTC (Tue) by BenHutchings (subscriber, #37955)
They don't even update kernels between releases.
Yes they do. And it's not just updating drivers; they've backported quite large new features, like KVM into RHEL 5.
Posted Apr 25, 2012 1:57 UTC (Wed) by jengelh (subscriber, #33263)
Any takers for (Solaris's) SMF? :)
Posted Apr 25, 2012 12:21 UTC (Wed) by jond (subscriber, #37669)
Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds