Closed mpoquet closed 7 years ago
Commit c5111bc restored the previous exec1 behaviour, which consisted in waiting the for socket to be usable before running the Batsim and Sched processes.
execN_ssh4 might still cause issues.
Exec1 and ExecN are now based on asyncio rather than execo. It may fix the issue.
Issue seems solved, closing the thread.
Since ZeroMQ commit d8a26cdae, tests seems to be not so deterministic. Various commits worked fine on my laptop, failed on the CI but worked when I retried to run them on the CI...
The problem might be in Batsim, in pybatsim or in the exec1 and execN experiment scripts. I suspect some TCP management should be done in execN before running the scheduler (making sure that the port is not being used).