Hi All,
Update on recent work done for the CentOS Container Pipeline.
- We
noticed that some test runs would fail due to content not being
available at the right time. We tracked this down to and fixed a race
condition that caused container test stages to remove content while it is being used in delivery stage. This also fixed a problem where sometimes the notify email was not being sent out.
- removing
all artifacts on the builder nodes once the built containers are pushed
out (or failed ), without this we were using up a lot of storage for
content that would otherwise never be useful. We will try and come up
with a smart process to catalog these before removing them ( either via
cron or opportunity driven )
- Refresh the build configurations and templates for each build, so that it builds from scratch for each build. This ensures that no previously cached content or state can impact the present build, test and delivery run.
- Due
to the layers involved with openshift and how we implement the build,
reporting cycle we've had issues with the emails that get sent out to
the user to notify them of a success / fail state along with some details
on what went wrong or where they can download the built images. We've
done some work in this area to both improve the content that gets sent
out and also to help ensure a better delivery rate ( we believe all the
issues causing email delivery failures are now fixed, but we will
continue to monitor )
Our immediate next focus is :
-
be able to demo some of the work we've been doing to the various SIGs
in CentOS, to make sure we are doing the best possible work to solve
their problems and support them in the best possible manner.
- continue to work on closing off the minimal feature set allowing us to release / announce the service for the wider audiences.
- develop a comprehensive life-cycle plan to audit and report on existing containers built in the pipeline.