Tools » teuthology

As a first pass, I think that we can create a table on teuthology.front.sepia.ceph.com with the following fields:

directory varchar(255)
pid       integer
result    enum(success, fail, not-run, hung, hung-with-error)
info      varchar(255)

directory would be the directory in /a corresponding to the suite.
pid would be the pid of a specific test.

result would be the result of the run:
    success -- passed (according to results.log)
    fail    -- failed (according to results.log)
    not-run -- hung (according to results.log) where no other information is available
    hung-with-error -- hung (according to results.log) where the error can be determined by looking at a left over teuthology.log or summary.yaml file
    hung    -- hung (according to results.log) where teuthology.log or summary.yaml indicates the test had not completed before the timeout)

info would contain text that can be gleaned from results.log (in the case of passes and failures) or the summary.yaml and teuthology.log files (in the case of hangs)

A new teuthology command, teuthology-update-db will be used to update the database. It would take the directory to use as input data to the database as a parameter.

A script can be written to populate the database with tests already in /a. After that, a call to teuthology-update-db can be added to schedule_suite.sh to make sure that the database stays up to date as tests are run.

Note that this table will be one flat table. After we get this set up, we can probably add other tables extracting index values out of the information in the directory field to be used as quick access keys if we wanted to spiffy things up.

couple thoughts:

- a separate column for the name of the run would be good so it doesn't have to be parsed out of the path
- and autoincrement id column, or should we make the path the primary key?