> Op 24 augustus 2016 om 6:38 schreef John Burwell <john.burw...@shapeblue.com>: > > > Wido, > > My only issue is dropping for any distro between patch releases. If someone > is running 4.9.0.0 on Ubuntu 12.04, and they need to update to 4.9.1.0+ (e.g. > to get a CVE fix), they will be stranded. This failure seems to fail the Law > of Least Surprise. While I recognize that it is unlikely that we have people > running 4.9 on 12.04, it is impossible to be certain. Therefore, I vote to > play it safe, and continue to support it in 4.9 release branch. >
I understand. That's why I said that Rohit should change the PR which is open now. However, what frustrates me a bit is that I submitted the PR in May this year and we are having this discussion now, again. That PR should have been merged a long time ago. > For master (i.e. 4.10.0.0), Wido makes a strong case for dropping Ubuntu > 12.04. If any users are using Ubuntu 12.04 when 4.10.0 is released, they > would have a supported release well past the April 2017 EOL since 4.9 is an > LTS release. Therefore, removing Ubuntu 12.04 support from 4.10.0.0 seems > like a Good Thing (tm) in terms of simplifying the code and testing matrix. > > Can everyone accept that the 4.9 release branch will be the last to support > Ubuntu 12.04? If so, we can repoint the PR and merge it. > > In terms of Ubuntu 16.04 support, ideally we would support it in 4.9.1.0+. > However, if I understand Wido correctly, supporting Ubuntu 12.04 and 16.04 in > the same branch is very difficult or impossible. Am I correct in my > understanding? > Well, I am not 100% sure what will happen. Didn't test it either. It will probably still work, but I can't guarantee it. Wido > Thanks, > -John > > > > john.burw...@shapeblue.com > www.shapeblue.com > 53 Chandos Place, Covent Garden, London VA WC2N 4HSUK > @shapeblue > > > > On Aug 23, 2016, at 6:14 AM, Wido den Hollander <w...@widodh.nl> wrote: > > > >> > >> Op 23 augustus 2016 om 11:38 schreef Rohit Yadav > >> <rohit.ya...@shapeblue.com>: > >> > >> > >> Historically, CloudStack's debian/deb packages/repositories have never > >> been supported by the initial authors. For example, initial ACS version > >> and all CCP releases never shipped deb packages, nor in our (old and > >> recent) documentation we promote installation/running CloudStack on > >> Debian/Ubuntu. Afaik, it was Wido who introduced it because he wanted to > >> run CloudStack on Ubuntu/Debian-based distro. Also, the packages are > >> something that the project never shipped or endorsed or supported, so it's > >> up to the maintainers of various repositories how they are building and > >> hosting CloudStack packages. Even if we remove the packaging support in > >> our branch/repository, anyone can build CloudStack for any distro, several > >> people/projects have packaging related buildsystem/code separated from the > >> project codebase. Most tutorials that I found are based around Ubuntu > >> 14.04 or CentOS, given that 12.04 is 4+ years old, we might not even have > >> anyone using CloudStack on it. > >> > > > > I highly doubt somebody still runs Ubuntu 12.04 with a recent version of > > CloudStack. > > > > 4+ years in Qemu/libvirt time is a very long time. > > > >> > >> That said -- I think 4.9 should at least not drop the support yet, just to > >> let any Ubuntu 12.04 user who may be using it in the wild. If we look at > >> the PR, the way we're dropping the support is by simply bumping up few > >> package dependency versions. The issue of supporting or dropping support > >> for Ubuntu 12.04 lies in those version changes only. > >> > >> > >> The more important thing right now is to support at least Ubuntu 16.04 > >> hosts as KVM guests and usage-server hosts, which is much needed in both > >> 4.9 and master branch for the upcoming 4.9.1.0 and 4.10.0.0 releases. > >> > >> > >> Wido -- would it be acceptable to avoid bumping up the min. package > >> dependency version, i.e we don't change the pkg dependencies for > >> cloudstack-agent and keep the version number as it is for lsb-base, > >> qemu-kvm, libvirt-bin for 4.9 branch. While on 4.10, we can discuss if we > >> want to drop the support now or plan this later. > >> > >> > > > > Well, yes. But I don't know *what* might break on 12.04. I wrote the PR in > > May and there must have been a reason for that. > > > > Feel free to modify the PR and not bump those versions. Packages might work > > or not, not completely sure. > > > > Wido > > > >> Regards. > >> > >> ________________________________ > >> From: Wido den Hollander <w...@widodh.nl> > >> Sent: 23 August 2016 11:38:43 > >> To: dev@cloudstack.apache.org; John Burwell; us...@cloudstack.apache.org > >> Subject: Re: Eliminating Support for Ubuntu 12.04 > >> > >> > >>> Op 23 augustus 2016 om 1:02 schreef John Burwell > >>> <john.burw...@shapeblue.com>: > >>> > >>> > >>> All, > >>> > >>> PR 1647 [1] proposes dropping support for Ubuntu 12.04 from 4.9.2.0+. > >>> The primary motivation for its removal is that the age of its libvirt and > >>> qemu versions greatly complicate maintenance of the KVM integration. > >>> However, Ubuntu 12.04 will be supported until April 2017 [2]. What would > >>> be the impact to our user community of removing support for Ubuntu 12.04 > >>> before its EOL in April 2017? If we don’t drop support for it in 4.x, > >>> would it be acceptable to drop support for it in 5.0.0 which is currently > >>> scheduled for release at the end of 2016 [3]? > >>> > >> > >> The PR was supposed to go into 4.9 already, it just took way to long to > >> get merged. So that's why it would now go into 4.9.2 > >> > >>> If we do chose to drop support for Ubuntu 12.04 in 4.x, I propose that we > >>> remove it in 4.10.0.0 rather than 4.9.2.0. First, it is reasonable for > >>> users to expect that upgrading between patch releases (i.e. 4.9.x.x -> > >>> 4.9.x+1.x) would not require system changes. Dropping a distribution > >>> would violate such an expectation. Second, 4.9 is an LTS branch. > >>> Therefore, maintaining 12.04 support in 4.9 would provide LTS users with > >>> support for Ubuntu 12.04 until May 2018 — well after its EOL. Does this > >>> approach seem reasonable if we elect drop Ubuntu 12.04 in 4.x? > >> > >> Again, this PR had to be merged earlier, not later. It's just very > >> difficult packaging wise to keep supporting 12.04 and 16.04 at the same > >> time. > >> > >> The Qemu and libvirt versions in 12.04 are truly ancient. I doubt anybody > >> is running stock 12.04 with CloudStack 4.8 right now for example. > >> > >> The Ubuntu 12.04 debate has been popping up multiple times in the last > >> year. > >> > >> This PR has been open way to long, that's imho the main problem here. > >> > >> I'm in favor of dropping 12.04, should have been done in 4.9 already and > >> not wait any longer. > >> > >> Wido > >> > >>> > >>> Thanks, > >>> -John > >>> > >>> [1]: https://github.com/apache/cloudstack/pull/1647 > >>> [2]: https://wiki.ubuntu.com/Releases > >>> [3]: > >>> https://cwiki.apache.org/confluence/display/CLOUDSTACK/%5BPROPOSAL%5D+2016-2017+Release+Cycle+and+Calendar<https://cwiki.apache.org/confluence/display/CLOUDSTACK/[PROPOSAL]+2016-2017+Release+Cycle+and+Calendar> > >>> > >>> > >>> > >>> john.burw...@shapeblue.com > >>> www.shapeblue.com<http://www.shapeblue.com> > >>> 53 Chandos Place, Covent Garden, London VA WC2N 4HSUK > >>> @shapeblue > >>> > >>> > >>> > >> > >> rohit.ya...@shapeblue.com > >> www.shapeblue.com > >> 53 Chandos Place, Covent Garden, London WC2N 4HSUK > >> @shapeblue >
