[Gluster-devel] Project Update: Containers-based distributed tests runner
avishwan at redhat.com
Fri Jun 14 14:13:25 UTC 2019
**gluster-tester** is a framework to run existing "*.t" test files in
parallel using containers.
Install and usage instructions are available in the following
- Create a base container image with all the dependencies installed.
- Create a tester container image with requested refspec(or latest
master) compiled and installed.
- SSH setup in containers required to test Geo-replication
- Take `--num-parallel` option and spawn the containers with ready
infra for running tests
- Split the tests based on the number of parallel jobs specified.
- Execute the tests in parallel in each container and watch for the
- Archive only failed tests(Optionally enable logs for successful tests
- NFS related tests are not running since the required changes are
pending while creating the container image. (To know the failures run
gluster-tester with `--include-nfs-tests` option)
- Filter support while running the tests(To enable/disable tests on the
- Some Loop based tests are failing(I think due to shared `/dev/loop*`)
- A few tests are timing out(Due to this overall test duration is more)
- Once tests are started, showing real-time status is pending(Now
status is checked in `<logdir>/regression-<job>.log` for example
- If the base image is not built before running tests, it gives an
error. Need to re-trigger the base container image step if not built.
- Creating an archive of core files
- Creating a single archive from all jobs/containers
- Testing `--ignore-from` feature to ignore the tests
- Improvements to the status output
- Cleanup(Stop test containers, and delete)
I opened an issue to collect the details of failed tests. I will
continue to update that issue as and when I capture failed tests in my
Feel free to suggest any feature improvements. Contributions are
More information about the Gluster-devel