There could be a corner case when an app was run on a cluster having varying number spark executor cores. We should be able to handle this case - correctly identify the source cluster shape and estimate the recommended cluster shape.
Example
A Spark App having executors with different number of cores: 96, 96, 16, 80, and 96
Note we saw this happen on an EMR cluster running Spark 2.x. So this was on yarn and I have never seen executors have different number of cores on YARN.
There could be a corner case when an app was run on a cluster having varying number spark executor cores. We should be able to handle this case - correctly identify the source cluster shape and estimate the recommended cluster shape.
Example
A Spark App having executors with different number of cores: 96, 96, 16, 80, and 96
cc: @tgravescs