Implementing a Process Manager

[GuiguiMtl]

On a previous Issue I asked of the difference between Process Manager and Session and ended up creating a process manager.
One of the issue I am facing is that as a result several Aggregate are being tracked in the Session. When I commit for the first time the tracked aggregates are not cleared therefore still tracked. That's probably because
So during the second commit, it tries to save the aggregate previously save again but this create a Concurrency Issue because the version of the aggregate descriptor is not updated.

One solution I have found it to update the commit with this
await Task.WhenAll(_trackedAggregates.Values.Where(aggregate => aggregate.Aggregate.GetUncommittedChanges().Any()).Select(x => _repository.Save(x.Aggregate, x.Version, cancellationToken)));

In order to only save Aggregates that does not have any other changes to save.

The other option would be to not wait for the save of aggregate to be finished to actually clear the tracked aggregates of the session but I can't figure a way to make that work.

Does that make sense or am I going to break something I do not realize yet ?

[gautema]

Hi. I’m on vacation, so I don’t get to try anything out, but I don’t get why the session isn’t cleared. Could you please elaborate? Could it work to create a new session for the second commit? Mvh Gaute Magnussen

…

On 26 Mar 2018, at 19:41, GuiguiMtl ***@***.***> wrote: On a previous Issue I asked of the difference between Process Manager and Session and ended up creating a process manager. One of the issue I am facing is that as a result several Aggregate are being tracked in the Session. When I commit for the first time the tracked aggregates are not cleared therefore still tracked. That's probably because So during the second commit, it tries to save the aggregate previously save again but this create a Concurrency Issue because the version of the aggregate descriptor is not updated. One solution I have found it to update the commit with this await Task.WhenAll(_trackedAggregates.Values.Where(aggregate => aggregate.Aggregate.GetUncommittedChanges().Any()).Select(x => _repository.Save(x.Aggregate, x.Version, cancellationToken))); In order to only save Aggregates that does not have any other changes to save. The other option would be to not wait for the save of aggregate to be finished to actually clear the tracked aggregates of the session but I can't figure a way to make that work. Does that make sense or am I going to break something I do not realize yet ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[GuiguiMtl]

Thanks for your answer.

The session is not cleared because after committing the first aggregate, the event is broadcasted to the system. It is handle by the event handler which creates and save the process manager. At this point the session keeps track of the aggregates that was saved previously and the process manager.

That's because even if it is asynchronous everything is done in this given order :
->Commit
-> Flush changes to aggregate
-> Save event to EventStore
-> Publish Events
-> Event Handler Handles Event
-> Launch Process Manager
-> Commit changes made on the process manager
-> Concurrency exception because we are still in the save of the first commit. As we did not reach the clear aggregates statement of the first Commit, we try to save a version of an Aggregate which is the previous version.

Creating a new session in the handler would probably fix the issue but I feel like it would break the idea of Session you were trying to bring with it, no ?

[gautema]

Sorry the late reply. I don't think you should use the same session. Hopefully you have fixed this by now.