Change runner host between steps of a single workflow?


I have a workflow with several steps (just normal…) I have noticed the following behavior several times on macOS runners. Sometimes, from one step to another, the workflow continues on another machine (a distinct host name).

Does it sound normal? Is it expected?

Why do I bother? In my build system, the binaries are built in one directory, the name of which is built from the architecture, host name, debug/release, etc. The idea is that the built is possible on a share volume between systems, typically several VM’s in my case. Each machine builds in a separate area in the same repo structure.

So, in one step, the project is built. The build directory is

This means that the host name is “Mac-1720”.

In the next step, the unitary tests are run. And the binaries are searched in

This means that the host name is now “Mac-1784”.

Of course, the binaries are not found and the test fails.

OK, now that I know the cause of my intermittent workflow failures, I can adapt.
But my question is: is this normal, expected, documented, where?
I have seen this on macOS runners, not Linux runners.


Thank you for reporting this issue . And sorry for this unexpexted experience.
I submitted an issue in virtual-environment repo, please continue track this problem here:

Thanks. You spotted that the problem was even worse: the host name changes during the execution of a step. I could have imagined that the VM or container could have been migrated between steps. But the host VM does not change, the name of the host is changing during the execution.

So, now it is tracked on

1 Like