This message was deleted.

Awesome! Thank you. I think the documentation was missing the .targets on it's example.

any advice on getting hundreds of targets to action the same steps, but not wait on each other to finish before they continue?

You can use the

parallelize

function

thanks, we already do that. Where we get tripped up is in 'everyone do A', then everyone that successfully did A, do B, and everyone that failed to do A, do C instead. But don't wait until you know who succeeded and who failed before starting. seems like we are missing something relatively obvious.

I'm not sure I understand. If you have nodes X, Y, Z and they have to run tasks A, B, C, do you want to ensure every node has finished A before any node starts B, or do you just want each node to independently run A then B then C, but stopping if that node fails?

wow, two months later. the problem with your proposal, is that task b will run on hosts where task a failed

That shouldn’t be the case. Within the block, an exception (caused by the failure of a on a target) will stop execution of that block and not run b and c. But the parallelize block runs separately for each target, so it shouldn’t affect other targets.

yeah. it's been a while since I looked at this specific codebase, so I forget which problems we run into, and when

I think it should work but I’m not sure. In particular I don’t know if it will work to have two different tasks running on the same target at once. I think so but I could imagine it not working.