Closed ipleten closed 4 years ago
Hmmm, yeah need to check - but at first looks possible.
Just returned from vacation - so can work on the issue. Seems like standalone object is not so useful during the pipeline execution - will check how to fix that.
Sorry, closed by mistake
So, result of the research (sorry for the delay - really busy) - the standalone objects state is not saved in the program.dat in the active build (which contains the CpsThreadGroup state). For now I see no resolution - because MPLManager is storing the state of the pipeline and right now I see no way to restore the MPLManager instance variable even from the existing MPL variable (which is saved in the program.dat)...
Need to think about it - but for sure it's a blocker for the 1.0.0 version...
I see 3 potential ways to resolve the issue:
MPLManager.instance
check the CPS nodes on existing object of MPLManager class - and use it to store into a static instance var.Hi @ipleten ,
Could you please check the provided fix griddynamics/mpl#55
Pruvit, @sparshev! We have quick tested changes and jobs are being resumed now. Thank you!
Great news! Thank you) Just merged the PR.
We have an
input
stage on lightweight worker which could be sitting very long until user input. So with job waiting for user input and after Jenkins restart if user confirms the input (press "Deploy") we get following error. If I moved outinput
from module to 'pipeline.groovy' it works, but I expect that some other long running stages might fail after Jenkins restart. It looks like after restart by some reason MPL loses track ofactiveModules
.We added some debug info so the lines number in stack trace are slightly different, error is happening in https://github.com/griddynamics/mpl/blob/master/vars/MPLModule.groovy#L100
Jenkinsfile
module
The error I got is :