I released taskotron-trigger 0.3.10 and libtaskotron 0.3.13 with execdb and artifacts support. They are both in f21 infra-testing repo and deployed on dev machine.

I'd like to note here, so it's not forgotten, that we need to keep in mind, when deploying to stg and prod that:

1. we need to remove "deployment_type == dev" conditionals in master.cfg and roles/taskotron/taskotron-client/templates/taskotron.yaml.j2
2. upgradepath that supports artifacts lives in develop branch (which is default branch) and playbook still works with master branch - so merge upgradepath develop into master and revert default branch there + do git symbolic-ref HEAD refs/heads/master on dev

One of the things that we were planning to do with the stg re-deploy was to get artifacts and execdb deployed, so let's get that done for stg.

Stg is also going to need some proxy changes for execdb but those should be pretty straight-forward

proxies configured for execdb in stg

FYI artifacts and execdb deployed on stg. We're still getting some errors though...

I updated taskotron-proxy so that the execdb frontend works: http://taskotron-dev.fedoraproject.org/execdb/

execdb was never actually deployed on stg. Since this requires creating an execdb_stg db on db-01.qa (which is a frozen system), I've submitted a freeze break request. Assuming that is acked, I'll finish up deploying execdb on stg so that trigger starts working again.

The freeze break was approved, so I installed execdb on resultsdb-stg01.qa

With the new libtaskotron package, all the jobs started failing due to root ownership of /var/tmp/taskotron. I manually removed the directory and filed this ticket so we can fix the issue for real.

Then all the tasks started failing because the new libtaskotron version is expecting a more recent build of resultsdb_api. I bumped the version there, did a release, did a build in copr and added the new resultsdb_api to the infra-testing repo.

The deployed version of resultsdb doesn't support UUIDs, so I did a release of that and resultsdb_frontend (not required but it has the search button fix, better 404 handling), moved them to infra-testing and deployed to dev. Found out that the alembic bits weren't included in the resultsdb package, so I fixed that and deployed so that the alembic upgrade would work and the db-accessing code wouldn't puke.

I think that both dev and stg are working now. Leaving this open until we're sure

@tflink Wow, thanks for taking care of this. Sounds like it was "fun" :/

Artifacts has been deployed on all machines.