The build and deployment of rest base services could be improved quite a bit.
Given the variety of configuration of OS systems we run it is likely to run into problems with interactions among OS and Docker while building. While builds could happen on a labs machine that is just a temporary fix. The build should happen on jenkins as part of the CI system. To build "aqs" sql lite could be used so there is no need for jenkins to run a cassandra instance.
There are other problems though: after a change is build it gets pushed to gerrit (a huge chageset with a bunch of binaries). This changeset is merged to master and deployed from there. A rollback means another merge (or unmerge) an operation in the depot regardless. This also seems suboptimal and subjected to problems due to lack of package versioning.
I understand that these are not problems that can be fixed right away but a services platform need to have a solid ecosystem and this build and deploy system seems rather brittle.
I am asking that the services team puts some work next quarter to improve the build and release process of restbase services.