rhettg / Tron

Next generation batch process scheduling and management
Other
11 stars 0 forks source link

Introduce concept of JobGroups ? #27

Closed mtai closed 13 years ago

mtai commented 13 years ago

It would be useful if Tron understood that a user may want to spin up a group of jobs all at once with the same type of scheduling or node pool.

JobGroup nodes: NODE_POOL ? NODE? scheduler: FALLBACK_SCHEDULER

Any job defined in the same job group would fall back to the entire group's "default" nodes and schedulers if they aren't provided on the job.

tronview would be able to print out just information on a "group" of workers. Imagine "tronview service_worker_group" would print out a basic tronview of JUST jobs that are in the same group. Then users can run tronview "certain_service_worker" to get job detail like normal.

tronctl would be able to run entire groups at a time. Imagine "tronctl start service_worker_group" that would go ahead and run all jobs in the same group.

rhettg commented 13 years ago

This description is really kinda confusing.

How would the user decide between Jobs and Actions or Job Groups and Jobs ?

I can definitely see a use for binding jobs together, mostly for display and reporting requirements. Like email all failures to a specific user or team. But probably NOT for controlling together or providing configuration defaults. That would just get too confusing.

mtai commented 13 years ago

I can see how this would be confusing. I think you get the gist of it though. Reporting/display were my main concerns here as a recent "tronview" run basically outputted a jumble of jobs that I didn't really care about. As we start to collapse more and more fcron entries into Tron, a "tronview" run by itself without a job name will basically become useless.

Maybe a job has a "notification_list" or "reporting_group" or some other concept that tronview hooks into? Leaving this up to you guys

mtai commented 13 years ago

Closing this issue out in favor of another suggestion