I was wondering how the clustering works, i read the documentation and it was stated that the token will be resumed on another node. So i did a test, i make a small workflow that has to print out the node which is handling it, and it gets and sets a configuration element, so that i could see that it resumes and not restarts it.
Each scripting task is adding +1 , +2, +3 to the corresponding conf element inside vro. So that first element if it is 1 , it means it ran just once, if it was 2, it would mean that the workflow was restarted not resumed. on 1st script i add 1 , on 2nd i add 2, on third i add 3 to firstvar,secondvar,thirdvar configuration element variables.
I put sleep 90s between each scripting task. It started off from node 761.
After couple of seconds from start, i have powered off vro node 761. Another node took the execution from there:
Configuration element variables were initially set all to 0. After finishing the run they were set to 1,2,3:
Each script task had just to ouptut date/time + the host that ran it + get the content of configuration element variable + set it + get it again.
So to sum it up, yes, the orchestator nodes in cluster pick up the workflow job to resume in case their node on which they were running will be unavailable.
When i have powered on back the failed node vro761 (761,762,763 -> cluster7) suddenly the Logs section was put all together, and you can see how the it was handled:
