- Rundeck

Date: 06 May 2013

Server: , Channel: #rundeck

19:21:01 travis-ci

[travis-ci] dtolabs/rundeck#235 (development - 90a25f3 : Greg Schueler): The build passed.

19:21:01 travis-ci

[travis-ci] Change view :…90a25f33060a

19:21:01 travis-ci

[travis-ci] Build details :

21:30:01 travis-ci

[travis-ci] dtolabs/rundeck#236 (development - cdfa1c8 : Greg Schueler): The build passed.

21:30:01 travis-ci

[travis-ci] Change view :…cdfa1c8963b8

21:30:01 travis-ci

[travis-ci] Build details :

22:08:33 kindjal

is it possible to have a workflow where step 1 applies to node set N, and step 2 applies to node set M?

22:15:16 silk

you can create 2 jobs, with different node sets, and wrap them in a job that calls each in turn

22:15:19 silk

if thats what you meant

22:15:37 kindjal

yeah I think I don’t know about job nesting

22:15:45 kindjal

what I want is

22:15:51 kindjal

issue a command

22:15:57 kindjal

then watch and wait for something to happen

22:16:03 kindjal

then when I see that thing happen run another command

22:16:09 kindjal

repeat that over sets of nodes in turn.

22:16:41 kindjal

I have an HPC cluster with 400 nodes. I want to close them to new jobs, wait for jobs to finish, then run a command, then re-open them, in batches of say 10 or 20 nodes.

22:18:53 silk

im not sure I completely understand, but you want to execute a cmd (close), wait for something, execute a cmd(open) on 10 nodes, then move to the next batch of 10?

22:19:11 silk

until all 400 nodes have had that happen?

22:26:02 kindjal


22:31:09 kindjal

seems like you have to define a target set of nodes with a job

22:31:19 kindjal

rather than defining a job, then applying it to a set of nodes in a workflow.