Rewind deployed processes

[edoshor]

Hi,

We've suffered from automatic re deployments of our processes (http://jira.codehaus.org/browse/ACT-736) for quite a while now.
We plan to fix that by migrating to 5.10.

We fear of slow queries due to tables loaded with needlessly duplicate and irrelevant process definitions.
Is it possible to "collapse" these re deployments ? any kind of manual manipulation of data in the DB ?

Our processes did not change for a long time as well.
We use full history - important for us.

[jbarrez]

How many process definitions are we talking about?

You can easily delete deployments through the repositoryService. When doing so, it will cascade to all involved data.

[edoshor]

We've 3 process definitions over all.

No. I don't want to delete any instances or history.
We've ~2500 historic process instances spread over ~100 re deployments of the same process definition.
We would like to do something in order to make them belong to the same, single process definition.

We don't expect anything like this in the API. We are willing to write some code to do this.
We're concerned if its even possible and if so what we have to do.

[jbarrez]

You could choose not to cascade when you delete a deployment. That keeps the history intact.

You will need some SQL magic queries to do it. Like with code, nothing is impossible. But the schema is not that difficult, and has no foreign keys, so it should be doable.