Passivation allows a requestor, service, or clipboard page to be saved into the PegaRULES database and reactivated later. A background process known as the passivation daemon converts the in-memory state and clipboard information into database rows, making the memory available for other uses. The reverse process is known as activation.
Passivation is controlled through the prconfig.xml file or Dynamic System Settings. See How to create or update a prconfig setting.
Passivation occurs through three distinct mechanisms:
Passivation frees up JVM memory, making more memory available to other requestors. In addition, Timeout passivation has security benefits, and HTTP passivation has failover benefits.
Note: Idle pages within an otherwise active requestor session may indicate a design or implementation issue that merits research. Ensure that your processing creates only pages that are needed, and removes them when they are no longer needed.
Passivation details can be saved either as disk files on each Pega 7 server, or as database rows in the PegaRULES database. Which option you choose determines which passivation features are available.
By default, passivation details are saved as files on each Pega 7 server. This corresponds to the following prconfig.xml file setting:
<env name="initialization/persistrequestor/storage"
value="file system" />
The saved information is converted to Java serialization file (SER file type) in a directory named PassivationData, a subdirectory of the temporary files directory. For performance reasons, files are distributed in up to 36 subdirectories of this directory.
On a large system, ensure that adequate disk space is available for these files. Optionally, you can use tactics such as disk striping and RAID over multiple disks for maximum performance. Through a symbolic link, you can place this directory on a dedicated high-performance storage solution.
It is not recommended to save passivation details to the database. See the High Availability Administration Guide for details about passivation to shared storage using custom passivation.
If you must save passivation details in the database rather than the server file system, add the following setting to the prconfig.xml file:
<env name="initialization/persistrequestor/storage" value="database" />
When this option is set, the passivation daemon saves requestor information as an instance of the System-Requestor-Context class, corresponding to the pr_sys_context
database table. The daemon saves page details as an instance of the System-SavedPages class in the pr_page_store
database table.
To recover database space, the system automatically purges pages saved through this mechanism after three days.
When a requestor creates a page but does not access it for a long period, the passivation daemon may passivate the page class, and remove it from the clipboard. Other pages for that requestor remain in memory, and requestor processing can continue normally until that page is needed (for read access or for update). The page can be reactivated into memory on-demand.
Similarly, when an entire Thread or requestor is inactive, the Thread context or requestor context can be passivated (including the associated clipboard pages).
The passivation daemon wakes once a minute to handle passivations. By default, the passivation daemon passivates:
Typically, page passivation improves performance for production systems that may have 50 or more simultaneously connected requestors (not counting agents). For smaller systems (with adequate JVM memory), the benefits may be small or none at all.
To disable page passivation, add the following element to the prconfig.xml file:
<env "initialization/persistrequestor/usepagelevelpassivation" value="false" />
When enabled (through Initialization/PersistRequestor
entries in the prconfig.xml file or Dynamic System Setting), each HTTP requestor session can be saved after a specific time period of no activity.
To enable timeout passivation, use the following prconfig.xml setting:
<env name="Initialization/PersistRequestor" value="OnTimeout"/>
To prevent timeout passivation, use the Never
value:
<env name="Initialization/PersistRequestor" value="Never"/>
If the PersistRequestor
element is not present, the system assumes OnTimeout
.
Note: The setting Initialization/NoAuthenticateOnActivate
defaults to true. As a best practice, do not override this default setting, since other mechanisms may cause users to be prompted for credentials.
In a multinode cluster (with appropriate configuration), requestor passivation can promote high availability and load balancing across nodes, since the reactivated session can resume on a different node than the node that previously supported the user. (The previous node may be down or busy.)
Passivation at HTTP interaction end should be set to "OnTimeout" and should not be changed for High Availability starting with Pega 7. See the High Availability Administration Guidefor details.
Passivation also frees JVM memory, making more memory available to other requestors.
Note: HTTP passivation to support failover in a cluster can occur only when passivation details are saved to the PegaRULES database, rather than the file system.
For IAC users who access a Pega 7 application through a web node, passivation occurs at HTTP interaction end occurs always. The value of the Initialization/PersistRequestor
setting does not affect these users.
Clipboard pages that remain unused for long periods may indicate a design or implementation flaw that can hurt performance; the application created a page but neglected to use the Page-Remove method to remove the page after it was no longer needed.
When building or testing an application, you can discover whether your own processing has created such "orphan" or "near-orphan" page:
To see a list of passivated requestors on the PegaRULES database, open and run the standard list view report System-Requestor-Context.RequestorContextList.ALL. This information is also available from the System Management Application:
Passivated users are not automatically required to reauthenticate upon activation.
When a user exits a requestor session by logging off, all user clipboard pages are erased, including those pages not explicitly saved to the PegaRULES database by the Obj-Save method and Commit method. No passivation occurs.
Agent requestors have a BATCH
requestor type. They are never passivated.