Provisioning Scripts vs Postponed Operations
MidPoint supports operation postponing, for example during resource maintenance, unavailability, or when grouping the operations. However, as of 4.8, this postponing does not involve execution of provisioning scripts.
See e.g. MID-8942.
The primary reason is that the scripts to be executed are simply not recorded in
They are computed by the
model-impl (just like the deltas) and then passed to
provisioning-impl for execution.
(They rely on the lens context, hence it is not possible to prepare their execution without it.)
There are also special
reconcile scripts that are executed outside any provisioning operation.
We don’t talk about these here.
They are not postpone-able at all.
The more general reason for this is that we are not sure how exactly should the script execution postponing should look like.
There are two kinds of scripts: "before" and "after" operation ones. Hence, in general, one can have a set of scripts to be executed before any operation, then there is the operation itself, and after it, another set of scripts to be executed.
There are a lot of questions:
Should we keep track of what scripts were executed successfully, and postpone only those that were not?
Is this safe and error-proof enough?
If the operation itself fails, what scripts should be re-executed?