Skip to main content

Notice: this Wiki will be going read only early in 2024 and edits will no longer be possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.

Jump to: navigation, search

TPTP-Test-Tools-Design-Review-162605-12042007


(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Date

  • December 4, 2007

Attendees

  • Present:
    • Paul Slauenwhite
    • Duwayne Morris
    • Kent Siefkes
    • Joe Toomey
    • David Chadwick

Background

  • AG review of enhancement 162605 (Test execution should support parallel execution of tests).

Minutes

Took longer than expected, I accidently hit the save button after the Wiki timeout and lost what I had written. I shall not want to do that again.

There are two functional aspects of this feature:

  • 1. Execute multiple tests concurrently from the same Host IDE session.
  • 2. Launch and execute a test on a large number of agent machines in parallel.

As Joe explained during the call, it was previously the case that launching concurrent multiple tests was supported in TPTP. For example, while running a manual test, the user could launch a JUnit test that would run in parallel. There were failures under certain conditions. As a result, a decision was made to lock out this functionality to improve product reliability. Scott Schneider documented was was done and the problems encountered in 106852.

As a part of this enhancement, the direction is to restore this functionality, write automated tests to guard against regression, and fix the defects that are causing problems/failures.

The second issue is the major force behind this enhancement. A consumming product requires TPTP support to launch perhaps 50 agent machines in parallel as opposed to the current implementation, which is in serial. The issue is that the combination of initializing the agent, starting the session JVM, deployment of needed files via TPTP file transfer service, and starting the test runner is all done sequentially per agent. As a result, a run using 30 agent machines may take perhaps 45 minutes to initialize before starting test execution. Discussion focused on reducing this case to hopefully less than 5 minutes through a parallel launch of agents.

After various elaborations and discussions, the team arrived at the following:

  • 1. Implement execution of multiple tests in parallel using TPTP and write automated tests for this functionality.
  • 2. During development of parallel launching of multiple agents, testing of the functionality must include implementation and testing using the consumming product in order to be complete.
  • 3. Development and implementation steps should be done largely consistent with Joe Toomey's feature * Description Document.
  • 4. Effort should be focused on parallel execution for this feature with no serious effort made to improve the efficiency of existing functions (except perhaps discretionary easy picks).

Action items

DuWayne, make it happen.

Back to the top