On 05/25/2016 08:54 AM, Greg Freemyer wrote:
On Tue, May 24, 2016 at 6:38 PM, Richard Brown RBrownCCB@opensuse.org wrote:
In another thread, Yamaban brainstormed the following kind of schedule for future Leap/SLE releases
On 25 May 2016 at 00:20, Yamaban email@example.com wrote:
In light of the LTS kernel the only thing would be to move the release date of Leap / SLE, e.g.
- Kernel in January / settling in Tumbleweed
- Leap first Alpha in late February / early March
- Leap first Beta mid/late April
- SLE (sp) first Alpha early May
- Leap freeze late May
- SLE (sp) first Beta mid June
- Leap last full patch pull-in / first RC mid/late June
- SLE (sp) freeze early July / or better mid August
- Leap release mid/late July
- SLE (sp) last full patch pull-in / first RC early / mid September
- SLE (sp) release early / mid October
I want to compare that to where we are right now with SLE 12 SP2 and Leap 42.2
- Kernel in January / settling in Tumbleweed
- SLE 12 SP2 Alphas in Q1 2016
- SLE 12 SP2 Beta started late May <- NOTE: Leap 42.2 Development
cannot start in earnest until SLE sources are available, and SLE sources cannot be available until Beta
- Leap 42.2 first Alpha TODAY <- THIS IS WHERE WE ARE :)
- Leap Alphas & SLE Betas over the next few months (aka 'Q2')
- Leap Betas & SLE RC's after that (aka 'Q3')
- SLE 12 SP2 Release in 'Q4'
- Leap RC's in late Sept/October
- Leap 42.2 Release in November
Benefits of this current approach over what Yamaban proposed (Leap & SLE interleaved)
- Leap release is still in November, which in 10 years of openSUSE
releases we've proven that releases at that time for year works for us
- Winter/Spring releases get too disrupted by Christmas, Summer
releases get disrupted by summer holidays. October/November is the time that works for us best. 2. There is a nice symmetry of what we are doing with Leap and what SUSE are doing with SLE. Our major milestones are aligned. Developers working on SLE are directly benefiting Leap and visa versa 3. This timeline has the possibility for stuff like SUSE QA engineers actually working on Leap when SLE development is wrapping up but Leap will be at the prime interesting time of RC's and GMC. 4. We'll avoid the situation we had with 42.1 (where Leap released a few weeks before SLE 12 SP1) and then we had a pretty huge 'mega update' to sync up Leap 42.1 with the SLE 12 SP1 shared packages 5. It accepts the reality that Leap is a fair bit bigger (in terms of package numbers) than SLE - I think Yamabans suggested timeline is a little too aggressive. Even if we could realistically start Leap development before our base system had even started to come together, trying to properly integrate our 8000+ source packages in that kind of timescale is pretty much impossible.
Downsides of the Yamaban proposed approach (the "Leap before SLE" schedule)
- Um..actually I'm no sure there is any benefit..a July release is
pretty much impossible when most of the community goes on vacation over summer..releasing that much earlier than SLE means we'll have to miss out on a whole bunch of stuff or have a rough few months keeping everything in sync as SLE comes together..yeah..um..sorry I WANTED to try and write a balanced approach but I think our current roadmap is the absolute best plan we can have given our desire to release something annually and with what SUSE are planning for SLE 12 SP2 --
I'm pretty much in agreement that accelerating 42.2 release schedule makes little sense..
But, we all know that one of the major issues for the whole Leap concept was the SLE kernel was too old, so it was decided with 42.1 to use the most recent LTS kernel at the time of release.
The newly decided kernel LTS release schedule is to do a LTS release early each year. This year it is the 4.4 kernel released Jan 10, 2016. It will be roughly 10 months old when Leap 42.2 is released.
I don't have a solution, but I personally find that contradictory to expectations for a Leap kernel.
My thoughts as to options are:
- delay the full Leap release until early year so the yearly LTS
kernel can be incorporated
I could live with that but I don't like it.
- Maintain the Nov 42.2 release date, but do a 42.2 "advanced kernel"
release 6 months after 42.2 is released that updates the kernel to the 2017 LTS kernel
---> Note that the 42.1 kernel is scheduled to go out of support 6 months after 42.2 is released
---> Note that if SLE is going to do a summer 2017 SP, then having the kernel released into Leap 3 to 6 months earlier will allow them to get useful wide scale testing in advance.
if the kernel team is willing to support 2 kernels simultaneously (as they have done for the last decade plus), then we could have:
- Every Fall do a a full Leap release, but based on the roughly 10
month old LTS kernel of the year. Have the fall Leap kernel supported for exactly 12 months after the fall Leap version is released.
- Every Spring do a "advanced Kernel" release based on the annual LTS
kernel. Support the "advanced kernel" for the last 12 months of that Leap version's support.
- During the last 6 months of each Leap version's support, only
support the Advance LTS kernel released in the spring after the initial release.
Nov 2016 - Release Leap 42.2 with 4.4 kernel
Spring 2017 - Release the Leap 42.2 "advanced kernel" (ie. the 2017 LTS kernel).
Fall 2017 - Release Leap 42.3 with the 2017 LTS kernel (no longer designated advanced).
Fall 2017 - Drop all support for the 4.4 kernel when Leap 42.3 is released
Spring 2018 - Release the 42.3 "advanced kernel" (ie. the 2018 LTS kernel)
Spring 2018 - Drop 2017 LTS kernel support (when Leap 42.2 support is dropped).
The end result is that over the summer 2 kernels are supported: the current LTS kernel and the previous year's LTS kernel.
But during the winter only one kernel is supported.
From the kernel team's perspective they may even be able to leverage SLE kernel support at all times because they will have to support SLE kernels longer than the 12 months I'm proposing. That assumes the SLE will be supporting each of the annual LTS kernels.
I guess the question to ask is how much do Leap users really care about which exact kernel version they use remembering most of the people that care about having the latest of everything are probably running tumbleweed. Personally as long as my hardware works I have no problem with a older kernel. On the other hand if it comes to April 2017 and theres little overhead in offering what will become the 42.3 kernel in 42.2 alongside the existing kernel we should probably think about offering it. In this case the bigger question is how to make it an optional update and how do we make people aware that they could install this update as we wouldn't want it to be default behavior.