2020-05-27 Integration Meeting Minutes
Date & Time
May, 27, 2020; 9pm PST
Agenda
Release
Infra
Docs
Misc
Minutes
Release:
Aluminium is now passing distribution-test because we removed lisp and netvirt from the managed distribution and we can see now
that CSIT for Al is not looking good. We will need to address this.#1 thing to address is the persistence that is failing. It may have to do with the new segmented-journal change. We have
an INTTEST ticket here to track.other change is the new 404 vs 409 data-missing change that will need to be addressed in CSIT code.
there is an API change dealing with Maps in project code, and we don't know how this will turn up in our CSIT jobs.
best approach would be to debug and fix all the controller jobs, which should help results in other projects' jobs.
Aluminium formal release is still not until September. Netvirt and Lisp are not getting CSIT visibility at all and all other failures
need to be debugged and fixed. This appears to be quite a bit of work.The normal schedule allows for CSIT to have some troubles, but needs to be fixed within 2 weeks of having a green distribution
passing. We only got that this week.Nothing to talk about with Magnesium.
Infra:
the netvirt project working on reducing job usage (cost reduction)
the following table will be used to reduce job count
ODL | Openstack | Weekly/Daily |
Aluminium | Stein | Daily |
Aluminium | Queens | Daily |
Aluminium | Rocky | Daily |
Magnesium | Stein | Weekly |
Magnesium | Rocky | Daily |
Magnesium | Queens | Weekly |
Sodium | Rocky | Weekly |
Sodium | Queens | Weekly |
We need to try to resize our instances in our jobs. The number one cost is on odl-high-cpu-8 which is an 8 core, 8GB RAM
which appears to be used in autorelease, netvirt and lisp. Maybe those jobs can be resized?Sodium is now done with it's final release, so we will move it's autorelease to weekly instead of nightly and also remove the
Java 11 version.Another small low hanging fruit is the fix the scripts so that robot is not trying to run when the controller is not even running.
@Srinivas Rachakonda has an idea to only trigger 3node netvirt jobs if it's accompanying 1node job passes 100%.
Docs:
A few small updates made today to clean up release notes.
Misc:
Action Items
@Anil Shashikumar Belur will publish the artifacts for sodium SR3
@JamO Luhrsen to send instructions for how to get notified for any new patches in the docs project
@Anil Shashikumar Belur can try to help @Srinivas Rachakonda get the netvirt jobs in order to the new plan