A side effect of the changes made in #72260 is that explanations for failures to assign persistent tasks do not mention nodes which are shutting down.
As an example, this is the assignment explanation from a cluster that contains 24 data/transform nodes that are shutting down and 21 master/ml/ingest nodes that are not shutting down:
The explanation seems bizarre in that it tells you that the transform is not going to be assigned to any of the non-transform nodes but does not tell you why it's not going to be assigned to any of the transform nodes.
We need to find a way to include explanations for why persistent tasks are not being assigned to nodes that are shutting down into the overall explanation of why a persistent task was not assigned.
A side effect of the changes made in #72260 is that explanations for failures to assign persistent tasks do not mention nodes which are shutting down.
As an example, this is the assignment explanation from a cluster that contains 24 data/transform nodes that are shutting down and 21 master/ml/ingest nodes that are not shutting down:
The explanation seems bizarre in that it tells you that the transform is not going to be assigned to any of the non-transform nodes but does not tell you why it's not going to be assigned to any of the transform nodes.
We need to find a way to include explanations for why persistent tasks are not being assigned to nodes that are shutting down into the overall explanation of why a persistent task was not assigned.