Dear Support,

I am currently trying to execute a workspace and I have problem with the following error:

Unable to create worker; parallelism level may be too high for the translation requested

Transformer 'pointFetcher': Unable to create worker; parallelism level may be too high for the translation requested

For a set of 15000 polygons, I try in parallel and in a custom transformer to read pointclouds in a external database and then compute statistics for those pointclouds. Still within the custom transformer and in parallel, the input polygon is written with its statistics as attributes. By doing so in parallel and in a custom transformer, I am not overwhelming my RAM. In fact, as soon as the statistics are computed, the variables containing the pointcloud and the statistics for a polygon are erased from the RAM.

When launching this process in a minimal parallel processing mode (4 parallel subprocess) and for 1000 polygons, I do not encounter any problem.

When I launch the workspace for the 15000 polygons, I am getting the upper mentioned error after reading ~2060 polygons.

I don't understand why the number of group should be a problem as 4 child process are the maximum allowed number of subprocess (minimal processing mode). If I understand correctly, the main program should send the most 4 jobs simultaneously and resend jobs as long as there are features to process.

This process has to work polygon by polygon, as aggregating them together will force me to read a huge pointcloud and thus saturating my RAM.

Is the parallel mode the right way to process these data. Why is FME complaining when the number of group is getting to high ?

In advance thanks for your helpful support,

Florian