47 lines
2.2 KiB
Plaintext
47 lines
2.2 KiB
Plaintext
I am announcing my candidacy for PTL for the Ironic team for the Ocata release
|
|
cycle. In case you don't know me, I'm dtantsur on IRC. I started working on
|
|
Ironic around late spring or summer 2014, and I'm probably best known as a
|
|
founder of ironic-inspector sub-project.
|
|
|
|
The Newton cycle was a huge breakthrough for the Bare metal project. Testing
|
|
upgrades in CI, multi-tenant networking, support for multiple compute services
|
|
out of box are just a few great and highly-expected things that have become the
|
|
reality during this cycle.
|
|
|
|
As for Ocata, I would like to concentrate on unfinished tasks, CI and
|
|
technical dept. To be more precise, if you happen to choose me as your new PTL
|
|
for the Ocata cycle, I would like to shift focus to and be the driving force
|
|
behind these fields of improvement:
|
|
|
|
* Driver improvements and unification
|
|
|
|
First of all, this comprises the driver composition reform - my long-term
|
|
commitment which I finally plan to fulfill this cycle. However, I would also
|
|
like us to think more about unification between drivers. Several non-core
|
|
things very from driver to driver: UEFI support, RAID support and
|
|
capabilities discovery immediately come to my mind as examples.
|
|
|
|
Finally, this involves parting with drivers that do not have 3rd party CI.
|
|
I am ready to share responsibility for this unpopular move :)
|
|
|
|
* CI improvements
|
|
|
|
The number of our jobs keeps growing, but we still don't cover a lot of
|
|
features that Ironic provides. We need to figure out the way to increase
|
|
coverage without increasing load on infra and number of transient failures.
|
|
A few approaches come to my mind. Using projects like TripleO as 3rdparty CI
|
|
may cover a few more things specific to its use case. Puppet experience with
|
|
"scenarios" may also come in handy to replace several jobs with one.
|
|
|
|
No matter which approach we take, we will need to make sure that it's still
|
|
clear (especially to newcomers) what caused a particular CI failure.
|
|
|
|
* Finishing Newton goals
|
|
|
|
We've done great job landing features in Newton, but a few important things
|
|
are still missing. For example, booting from volume, rescue mode, port
|
|
groups, rolling upgrades, dealing with automatic maintenance in a more robust
|
|
way, and some more.
|
|
|
|
-- Dmitry Tantsur
|