Mender Roadmap Yocto Integration

Hi,

please tell us the roadmap to support the latest and future releases of Yocto (e.g. dunfell, the new LTS)?

Mauro

1 Like

+1

Looking forward to Dunfell support and to an official policy addressing what Yocto’s new LTS release cycle means for Mender. I know this was somehow answered here but a stronger commitment to faster tracking of Yocto releases — at least LTS — would be warmly welcomed. Mender is now 2 major releases behind.

1 Like

Hi there!

Mender generally have followed the latest Yocto release in the following Mender release. The 2019 fall release (zeus) was an exception because there were so many changes in Yocto that caused problems for Mender (e.g. lsb removal) and required significant rework. However zeus support in meta-mender is right around the corner.

Like everyone we do have constraints on how much time we can invest in different areas like Yocto support. So although it would be really nice to support every Yocto release immediately and keep supporting it forever, this is not possible in practice. I know major board vendors support 1-2 year old releases but never the newest one, for example. Some only support the fall releases, not the spring ones.

With Yocto LTS this might change though. Would you rather Mender support the LTS for 2 years and not add support for the other yocto releases? What would be your preference?

Hi @eystein!

For our part, if we have to choose: LTS only.

We plan to stop upgrading to each new Yocto Stable release, and instead only upgrade to LTS releases. For us, the main rationale for upgrading Yocto/Poky/oe-core in the first place is getting security upgrades. The Yocto Project announced that oe-core would be covered by the 2yr support window, so we hope staying on the LTS release cycle will allow us to:

  • reduce the amount of dangerous, non-reversible upgrades we have to do (e.g. raspberry boot files which are often necessary for major Yacto upgrades, or the mender thud->warrior configuration files migration)
  • opt out of having to reinvest in Yocto every 6 months to stay on the security upgrade bandwagon
  • focus more on application-level upgrades instead of system-level upgrades
1 Like

Since zeus branch is still under test and Dunfell is already out as stable release, probably mender should skip zeus to move directly with a stable release for Dunfell.

I’m agree with @guillaumekh about the support for Dunfell, because it’s a very important release.

However, in my opinion 6 months is too short, but 2 years are too much. Probably it’s better to have something in the middle, like every year.
In this way it will be easier also to get latest versions of linux packages with the official support. For instance NodeJS has a new LTS every year and it is released in open embedded every year. I suppose that the same will happen for other softwares.

Focusing on application level upgrades it’s important, but sometimes these app upgrades depends by other linux software. Developing an app with possibly a restriction of 2-years old dependency it isn’t so good. Obviously, it depends by many factors.

1 Like

Hi Everyone,

Two months after my first post does it make sense to start integrating meta-mender into Dunfell based projects? can we consider meta-mender sufficiently stable on Dunfell, or stable enough to start integrating it into our projects?

If anyone in the community has some experience to share, I’d love to hear it.

Best,
Mauro

Meta-mender/dunfell does not work at the moment, but it is being worked on. Fortunately it seems to be easier to integrate with than zeus, which took a very long time, but some work is required. I cannot give an ETA at this point though.

Thank you @kacf, we hope the porting on Dunfell arrives soon :wink:

Best,
Mauro

1 Like

FYI, this is being tracked here https://tracker.mender.io/browse/MEN-3602