Currently, the shell script that is run for the CI jobs is defined in the jenkins/jobs/builders.yml. This makes it difficult to change what is run in CI, since it requires pushing the jenkins change to see the result. The only way to test in advance what a change to the shell script will look like is to manually create POC jobs.
Alternatively, we can create a top level ci-scripts directory with the following scripts:
get-node.sh - Responsible for getting a test node and setting the VIRTHOST env variable
collect-logs.sh - Responsible for collecting and archiving the logs
return-node.sh - Responsible for returning the test node
Then the builders would be simplified to:
bash get-node.sh
bash quickstart.sh <args for specific builder> $VIRTHOST
bash collect-logs.sh
bash return-node.sh
This would then allow us to easily iterate on simplifying the process to get and return a node to CI. It would be good to move the log collection from the khaleesi tree to its own ansible role, either in the oooq tree, or its own ansible-role-tripleo-log-collect repo.
Currently, the shell script that is run for the CI jobs is defined in the jenkins/jobs/builders.yml. This makes it difficult to change what is run in CI, since it requires pushing the jenkins change to see the result. The only way to test in advance what a change to the shell script will look like is to manually create POC jobs.
Alternatively, we can create a top level ci-scripts directory with the following scripts:
Then the builders would be simplified to:
This would then allow us to easily iterate on simplifying the process to get and return a node to CI. It would be good to move the log collection from the khaleesi tree to its own ansible role, either in the oooq tree, or its own ansible-role-tripleo-log-collect repo.