When running a model with CUBE Cluster Multi-Step implemented incorrectly the situation below could happen:

• you are running some CUBE Voyager scripts in a cluster node as step #1 -> generating, for example a matrix file called "step1.MAT"
• you are trying to use the matrix from the step above ("step1.MAT") in another cluster node #2 before the previous cluster node has closed the file (i.e., it has not finished running the script that was generating the file)

A representation of the problem (incorrect set-up) is provided in the screenshot below:

In this case, the step #2 will generate a warning message below in the Print File (note: this is just an example):

W(103): Error reading file step1.MAT @ I=372, M=1 IPzone=3000.

The meaning of this warning message is that the program was able to read the "step1.MAT" matrix file only up to zone 372.

The reason why this happens is the incorrect set-up of CUBE Cluster Multi-Step within the model above. In particular:

• Steps running in parallel should be completely independent between each other - not using files produced by the parallel step
• Wait4Files should be properly used, with .end cluster files, to make sure that the execution of a previous step is completed before starting the execution of a subsequent step that is dependent on the previous one.

If Cluster is not properly set-up, as described above, it can happen that "by chance" the previous step finishes running before the subsequent step start, and the model run till the end with no warning (or maybe without any failure in subsequent steps). But then, because of some runtime differences, it could happen that the problem arises apparently randomly. Therefore, Cluster needs to be set-up with independent steps and correct usage of Wait4Files, so that this kind of situations is avoided.