Hello! Here is a summary of what the team did from October 6 to October 19.


Regarding Software Factory:

- We did a lot of testing/fixing around deploying "pure" zuulv3
architectures (~20 patches on sf-config, packaging and so on) A list
of the problems we encountered and fixed can be found here:
https://softwarefactory-project.io/etherpad/p/zuulv3-issues-list
- We validated that nodepool2 (since we moved to rdo-pike) is still functionnal
- We are validating that zuul2/zuul3/nodepool2/nodepool3 are all
working together w/o any issue (needed for the migration)
- We use a new tool virt-customize, to build our image with the same
workflow used on the ci (install from packages instead of using dib
element)
- We improved our logserver retention policy that missed a lot of
files (prod + SF master)
- We experienced issues with rdo-cloud (fip / ceph ) so we did some
prod maintenance tool


Regarding RDO:

- logs.rdoproject.org was nearing a storage threshold, we started
implementing a 60 day retention policy.
- We've identified the root cause of the RDO Cloud I/O performance to
be the cache policy on the RAID cards


Regarding openstack-infra:

- Zuul v3 is officially in production once more and doing a rollback
to v2 is no longer considered
- Different improvements to the testing coverage of core base jobs in
order to prevent gate-wide regressions
- Troubleshooting of logs.openstack.org outage (disk space and inode exhaustion)
- Congratulations to David Moreau Simard, now a core contributor for
the project-config, openstack-zuul-jobs and zuul-jobs repositories
http://lists.openstack.org/pipermail/openstack-infra/2017-October/005613.html
- David and Paul will be hosting a "Zuul v3 ask me anything" for the
TripleO and RDO developers
http://lists.openstack.org/pipermail/openstack-dev/2017-October/123668.html


Regarding DLRN:

- We kept on cleaning up the Fedora Rawhide worker. Now less than 10
packages are failing to build.
- We have implemented a new (configurable) scheme for package release
numbering, which is more compatible with the Fedora version guidelines
and friendlier with rdopkg
- We have fixed a bug that prevented the --order command-line switch
from working
Regarding ARA:
- We worked on a WSGI middleware in order to load the web application
off of a logged sqlite database on the fly
https://review.openstack.org/#/c/511992/
- Ansible 2.4.1 was delayed, still waiting on its release in order to
ship a bugfix for ARA to work with 2.4
https://storyboard.openstack.org/#!/story/2001219


Regarding RDOPKG:

- We created a github2taiga importer for syncing rdopkg Issues with taiga.
- We wrote feature test scenarios for missing git tags and such in
preparations for better error messages.
- We debugged flaky gate results and test scenarios.
- We worked on commit message related RFEs.


Regarding RepoXplorer:

- All features have been made available via the REST API (it was
partial before). REST API has been documented
Regarding other things:
- David Moreau Simard and Paul Belanger are attending OpenStack Canada
days in Ottawa, Paul will be talking about Zuul v3
https://www.openstackcanada.com/
- The Zuul v3 migration has been resumed upstream. This has required
some work to adapt and fix our CI jobs for Packstack and TripleO.
The meeting can be watched at https://bluejeans.com/s/B2Dl3/


This sprint's taskboard is archived here:
https://tree.taiga.io/project/morucci-software-factory/taskboard/2017-42


The next public review will happen Thursday Nov 02 at 13:00 UTC, feel
free to join us on #softwarefactory for the event details.

The SF team

_______________________________________________
Softwarefactory-dev mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/softwarefactory-dev

Reply via email to