Chapter 14. Persistence

Binary persistence, or marshaling, converts the state of the process instance into a binary data set. Binary persistence is a mechanism used to store and retrieve information persistently. The same mechanism is also applied to the session state and work item states.

When you enable persistence of a process instance:

  • Red Hat Decision Manager transforms the process instance information into binary data. Custom serialization is used instead of Java serialization for performance reasons.
  • The binary data is stored together with other process instance metadata, such as process instance ID, process ID, and the process start date.

The session can also store other forms of state, such as the state of timer jobs, or data required for business rules evaluation. Session state is stored separately as a binary data set along with the ID of the session and metadata. You can restore the session state by reloading a session with given ID. Use ksession.getId() to get the session ID.

Red Hat Decision Manager will persist the following when persistence is configured:

  • Session state: This includes the session ID, date of last modification, the session data that business rules would need for evaluation, state of timer jobs.
  • Process instance state: This includes the process instance ID, process ID, date of last modification, date of last read access, process instance start date, runtime data (the execution status including the node being executed, variable values, and other process instance data) and the event types.
  • Work item runtime state: This includes the work item ID, creation date, name, process instance ID, and the work item state itself.

Based on the persisted data, you can restore the state of execution of all running process instances in case of failure or to temporarily remove running instances from memory and restore them later.

14.1. Configuring safe points

To allow persistence, add the jbpm-persistence JAR files to the classpath of your application and configure the decision engine to use persistence. The decision engine automatically stores the runtime state in the storage when the decision engine reaches a safe point.

Safe points are points where the process instance has paused. When a process instance invocation reaches a safe point in the decision engine, the decision engine stores any changes to the process instance as a snapshot of the process runtime data. However, when a process instance is completed, the persisted snapshot of process instance runtime data is automatically deleted.

If a failure occurs and you need to restore the decision engine runtime from the storage, the process instances are automatically restored and their execution resumes so there is no need to reload and trigger the process instances manually.

The runtime persistence data is to be considered internal to the decision engine. You should not access persisted runtime data or modify them directly as this might have unexpected side effects.

For more information about the current execution state, refer to the history log. Query the database for runtime data only if absolutely necessary.

14.2. Session persistence entities

Sessions are persisted as SessionInfo entities. These persist the state of the runtime KIE session, and store the following data:

Table 14.1. SessionInfo

FieldDescriptionNullable

id

The primary key.

NOT NULL

lastModificationDate

The last time that entity was saved to a database.

 

rulesByteArray

The state of a session.

NOT NULL

startDate

The session start time.

 

OPTLOCK

A version field containing a lock value.

 

14.3. Process instance persistence entities

Process instances are persisted as ProcessInstanceInfo entities, which persist the state of a process instance on runtime and store the following data:

Table 14.2. ProcessInstanceInfo

FieldDescriptionNullable

instanceId

The primary key.

NOT NULL

lastModificationDate

The last time that the entity was saved to a database.

 

lastReadDate

The last time that the entity was retrieved from the database.

 

processId

The ID of the process.

 

processInstanceByteArray

The state of a process instance in form of a binary data set.

NOT NULL

startDate

The start time of the process.

 

state

An integer representing the state of a process instance.

NOT NULL

OPTLOCK

A version field containing a lock value.

 

ProcessInstanceInfo has a 1:N relationship to the EventTypes entity.

The EventTypes entity contains the following data:

Table 14.3. EventTypes

FieldDescriptionNullable

instanceId

A reference to the ProcessInstanceInfo primary key and foreign key constraint on this column.

NOT NULL

element

A finished event in the process.

 

14.4. Work item persistence entities

Work items are persisted as workiteminfo entities, which persist the state of the particular work item instance on runtime and store the following data:

Table 14.4. WorkItemInfo

FieldDescriptionNullable

workItemId

The primary key.

NOT NULL

name

The name of the work item.

 

processInstanceId

The (primary key) ID of the process. There is no foreign key constraint on this field.

NOT NULL

state

The state of a work item.

NOT NULL

OPTLOCK

A version field containing a lock value.

 

workitembytearay

The work item state in as a binary data set.

NOT NULL

14.5. Correlation key entities

The CorrelationKeyInfo entity contains information about the correlation key assigned to the given process instance. This table is optional. Use it only when you require correlation capabilities.

Table 14.5. CorrelationKeyInfo

FieldDescriptionNullable

keyId

The primary key.

NOT NULL

name

The assigned name of the correlation key.

 

processInstanceId

The ID of the process instance which is assigned to the correlation key.

NOT NULL

OPTLOCK

A version field containing a lock value.

 

The CorrelationPropertyInfo entity contains information about correlation properties for a correlation key assigned the process instance.

Table 14.6. CorrelationPropertyInfo

FieldDescriptionNullable

propertyId

The primary key.

NOT NULL

name

The name of the property.

 

value

The value of the property.

NOT NULL

OPTLOCK

A version field containing a lock value.

 

correlationKey_keyId

A foreign key mapped to the correlation key.

NOT NULL

14.6. Context mapping entity

The ContextMappingInfo entity contains information about the contextual information mapped to a KieSession. This is an internal part of RuntimeManager and can be considered optional when RuntimeManager is not used.

Table 14.7. ContextMappingInfo

FieldDescriptionNullable

mappingId

The primary key.

NOT NULL

CONTEXT_ID

The context identifier.

NOT NULL

KSESSION_ID

The KieSession identifier.

NOT NULL

OPTLOCK

A version field containing a lock value.

 

OWNER_ID

Holds the identifier of the deployment unit that the given mapping is associated with

 

14.7. Pessimistic Locking Support

The default locking mechanism for persistence of processes is optimistic. With multi-thread high concurrency to the same process instance, this locking strategy can result in bad performance.

This can be changed at runtime to allow the user to set locking on a per process basis and to allow it to be pessimistic (the change can be made at a per KIE Session level or Runtime Manager level as well and not just at the process level).

To set a process to use pessimistic locking, use the following configuration in the runtime environment: 

import org.kie.api.runtime.Environment;
import org.kie.api.runtime.EnvironmentName;
import org.kie.api.runtime.manager.RuntimeManager;
import org.kie.api.runtime.manager.RuntimeManagerFactory;

...

env.set(EnvironmentName.USE_PESSIMISTIC_LOCKING, true);  1

RuntimeManager manager = RuntimeManagerFactory.Factory.get().newPerRequestRuntimeManager(environment);  2
1
env is an instance of org.kie.api.runtime.Environment.
2
Create your Runtime Manager by using this environment.