Bug 1035191

Ivo, 

Could you attach or link to a code snippet that shows how you create and send the CompleteTaskCommand command?

Thanks!

Created attachment 832651 [details]
server log

Fixed:
6.0.x: 
https://github.com/droolsjbpm/droolsjbpm-integration/commit/003e401b7bb4c94bb2e

master: 
https://github.com/droolsjbpm/droolsjbpm-integration/commit/9dfc47c60f65408c6d0

Hi Marco,

I do not think the client should be forced to request CompleteTaskCommand with deploymentId (JaxbCommandsRequest req = new JaxbCommandsRequest(deploymentId, command);). In REST it is possible to complete a task without the deploymentId and AFAIK that information is get internally https://github.com/droolsjbpm/droolsjbpm-integration/blob/master/kie-remote/kie-services-remote/src/main/java/org/kie/services/remote/rest/TaskResource.java#L242

Hi Ivo, 

The choice was between performance and convience: by making the deploymentId mandatory for these task commands, that means that the Task instance doesn't need to be retrieved, which is necessary in order to avoid race conditions (under load) when executing these commands:  

  CompleteTaskCommand
  ExitTaskCommand
  FailTaskCommand
  SkipTaskCommand

However, if convienence is more important than performance, I can change it. 

Are you sure about this?

Hi Marco,

Performance is the key - I believe that we all agree that the statement is true. If making the 'deploymentId' mandatory helps performance then we should not change it.

One thing I don't understand is how does the user provided 'deploymentId' remove the need to retrieve the task? Could you please provide a link to the source code? Or provide more context?

Another thing which is not clear to me is how can a user start/complete/etc. a task (using the remoting interface) which is not associated with a deployment? The business-central allows to create and assign tasks which are not tight to a process and therefor not bound to a deployment.

Many thanks.

@MB

Hi Marek, 

See the following commits: 

This shows the code as it is, where the deployment id is required: 

https://github.com/droolsjbpm/droolsjbpm-integration/blob/6.0.x/kie-remote/kie-services-remote/src/main/java/org/kie/services/remote/jms/RequestMessageBean.java#L361

You'll see that for task commands that are _not_ in the  "TASK_COMMANDS_THAT_INFLUENCE_KIESESSION" set of Task commands, no deployment Id is required, but if it is a command that influences the KieSession, it is required. 

For those small set of commands that do (see my comment above), the deployment Id is required in order to be able to (basically) lock around the kiesession. 

. 

Now, if we don't require the deployment id, then we _can_ get it, but only by first retrieving the Task instance. I actually coded up a solution for this, you can find it here: 

https://github.com/droolsjbpm/droolsjbpm-integration/blob/700e9fb0311d4feac544e4104ed345fadebfcb91/kie-remote/kie-services-remote/src/main/java/org/kie/services/remote/jms/RequestMessageBean.java#L354

This is the same "try" block as above, except that we're now doing the following: 

- If it's a TASK_COMMANDS_THAT_INFLUENCE_KIESESSION Task command, then we first retrieve the task. This is because the Task instance contains the deployment id information. 

This extra step is what will decrease performance. 

. 

Let me know what you guys decide -- the "solution" is available, if we choose to do that. I can always unrevert the original commit. ;)

Marek, with regards to your second question: 

"Another thing which is not clear to me is how can a user start/complete/etc. a task (using the remoting interface) which is not associated with a deployment?"

The majority of the Task operations are possible without submitting a deployment id. In fact, all Task operations that do not involve the process engine can be done without having to get the deployment id. 

However, the following commands all impact the process engine: they end up signalling the process engine that the associate _process instance_ should be signalled (that the task is complete and that the process instance can continue):

  CompleteTaskCommand
  ExitTaskCommand
  FailTaskCommand
  SkipTaskCommand

Because these commands affect the process instance, it's important to make sure that nothing else is concurrently doing something with the process instance, which is why we need the deployment id to lock around the kiesession that holds the process instance. 

Does that make sense? Please feel free to ask more questions if it's not!

At this time the task commands above require the deployment Id, This should be documented for GA1.

Vikram,

a workaround exists. A developer use this complete command:

Command<?> cmd = new CompleteWorkItemCommand(workItemId, results);

then he/she can use:

JaxbCommandsRequest req = new JaxbCommandsRequest(deploymentId, cmd); // here the deploymentId must be put there, this is currently mandatory for the commands stated above

instead of:

JaxbCommandsRequest req = new JaxbCommandsRequest(cmd);


The developer can get deploymentId information from TaskData (through GetTaskCommand), so the taskId is the only thing needed to complete a task, just he has to do this approach himself.

Also it is not a REST call as you stated but JMS.

Fixed. Commits: 

6.0.x: 
https://github.com/droolsjbpm/droolsjbpm-integration/commit/a8748cf0

master: 
https://github.com/droolsjbpm/droolsjbpm-integration/commit/d7aee7c9

This no longer has a target release of 6.0.0.

Verified in BPMS 6.0.1.ER1