Date: Thursday, 20 October 2011
Time: 10:00AM Eastern US (contact MichaelFiedler if you'd like to participate)
The Automation meetings alternate times each meeting to accomodate the global team.
Agenda
* Reoccurring agenda items:
* Main agenda items:
- Scenario ownership proposal - all are still available.
- Move each scenario to its own Wiki page
- Workgroup owner for each scenario
- Owners begin providing initial details for workgroup discussion
- Continue scenario review. See Automation Scenarios
- Manual automation execution
- Defining workflows across OSLC providers
- Others, as time allows
- Automation result contributions
- Automation Tool/Automation Provider interaction
- Previous Action Items:
- Next meetings:
- 27 October
- Continue discussion of manual execution scenario details.
Minutes
Attending: Michael Fiedler, Eric Bordeau, Brent Albrecht, Pramod Chandoria, Amy Wan, Robert Elves, John Arwe, Charles Rankin, Paul
McMahan? , Gili Mendel, Sheehan Anderson, Daniel Berg
Let
MichaelFiedler know if I missed anyone.
- Scenario Owners
- Mylyn Scenario : Tasktop (Robert Elves/Lucas Panjer) - will find a generic name for it
- Automation Provider/Automation Tool Interaction : Pramod Chandoria
- Manual Execution - Michael Fiedler
- Automation Result Contributions/Update - TBD
- Defining workflows across OSLC Automation providers - TBD
- Polling/Notifications - TBD
- Primary discussion centered around the Manual Execution scenario. Key discussion points
- Is the Automation Request artifact discussed in earlier meetings really required?
- Agreement after discussion was that it was not required - creation of an Automation Job in some "initial/created" state to represent both the Request and Result concepts is sufficient.
- Discussion on differences between tools with respect to what Automation Job creation means. Some test scenarios were discussed where it might not make complete sense, needs validation (see below).
- Discussion on the responsibility for and ability to update Automation Job by all actors who might need to.
-
- Discussion on the nature of Automation Job updates and contributions if this approach is taken
- OSLC should specify the attributes for the Automation Job and the mechanism for update (http PUT). Whether workers update directly or the automation provider update on their behalf is a tool decision.
- Examples of types of updates while execution is in progress could be updates to a status field, log file references, intermediate execution output, etc.
-
- Discussion on queuing of Automation Jobs in a "not-started" state
- Need for automation providers to wait for a qualified workers to execute the job.
- Some automation providers may just choose to fail immediately instead of queueing
- Also notion of "run on any available worker" or let the automation provider select the worker
- OSLC should not specify requirements for queueing or lack of queueing. Job attributes/states need to be flexible enough to allow for it.
-
- Discussion on Automation Job creation
- How are parameters/environment info/other input passed to the Job on creation?
- How do creators of Jobs know what input attributes to provide? Likely very tool specific
- Possible approach discussed where the provider exposes an Automation Job creation shapes artifact to document what it expects
- Discussion ended here with the agreement on next steps:
- Michael Fiedler will try to add details to this scenario and send it to the mailing list. Document the full lifecycle of an automation execution.
- We will have another meeting next week to discuss:
- Modifications/Corrections to the scenario
- Apply the scenario to different domains workgroup members have expertise in (build, deploy, test, DevOps? , automation tracking, etc) and look for holes.
Next meeting: Thursday, 27 October at 1PM Eastern US time