SR-D55106 · Issue 525464
Improvements to error recovery and logging for BIX extracts
Resolved in Pega Version 8.4
Previously, the BIX incremental extraction would restart from the last extraction date time if it had a failure, even if some batches were processed successfully, and the starting and ending datetime considered for the incremental extraction were not logged. As part of this update, extractions will resume from the point of failure instead of starting all over again, and the logging has been enhanced to include Information about the parameters used for extraction (starting time, end time, #of records processed/failed etc). However, if the -f flag is set, batch level failures will cause BIX to terminate but blob read errors will not. This causes the restart logic in this fix to not function properly in the face of blob read errors such as an ORA-01555 when reading the blob. Resolution for that aspect of this issue will be provided in a future release.
SR-D55161 · Issue 527652
getStoreDataTable updated to resolve thread lock issues
Resolved in Pega Version 8.4
An IlegalMonitorStateException in TableInformationMapImpl.getDataStoreTableInfo caused a read lock to not be released, preventing other threads from acquiring a ReentrantReadWriteLock.WriteLock on the same ReentrantReadWriteLock. This eventually lead to a Deadlock scenario forcing other threads into an irrecoverable WAITING state. The issue could be cleared by restarting the nodes, but the implementation of the getDataStoreTableInfo method has been updated to resolve this issue.
SR-D64220 · Issue 535133
Reinitializing Full Text Search will shut down Elastic Search regardless of search initialization status
Resolved in Pega Version 8.4
Pega search was only sporadically working after converting from standard sandbox-marketing to largesandbox-marketing. This was traced to the Full-Text Search initialization having failed on the app-tier node during re-initialization. As part of re-initialization, the existing node is shut down and the Elastic Search node is started. The shutdown process relied on a boolean flag which indicated if the current status of full-text search initialization was successful. In this case, ES was trying to validate all the cluster level settings, for every save of one of the cluster level settings, but could not due to topology changes in the Util Tier node. The boolean flag indicated initialization had failed so the shutdown process was not invoked, yet the system was attempting to start the second instance of an ES node on the same machine. To resolve this, the shutdown FTS code has been modified to shutdown ES regardless of the search initialization status.
SR-D65944 · Issue 533395
Parent case lock properly released when child case is resolved
Resolved in Pega Version 8.4
After creating Parent-Child casetypes with default locking where the child case had the “Allow access to parent” check box checked, the temporary lock acquired on the parent during resolution of the child case was not released afterwards. If “Allow access to parent” was not checked, then the locks were released on both the parent and the child. This was traced to a combination of parameters used by the openIfStale() API where aUnlockOnCommit could be set to false despite the provided locking strategy expecting it to be true, as well as honoring UnlockOnCommit when maintainLockingStrategy is false. To resolve this, the system has been updated to always check whether the lock is available in the map already and if it is, then set unlockonCommit to true. Otherwise, under all cases, honor the passed-in unlockOnCommit value.
SR-D16934 · Issue 493705
External Cassandra nodes listed in DDS cluster
Resolved in Pega Version 8.4
Nodes of an external Cassandra cluster were not listed in the DDS cluster except for the first one in the host list, and when the “only” listed Cassandra node was restarted, the status on the DDS cluster LP did not become “NORMAL” afterwards. In addition, even though other C* nodes were up and running, the external Cassandra cluster was reported as unreachable. This was an unintended side effect of work done on the landing page to reflect the real state of the nodes after some were killed and restarted, and has been corrected by refining the equals() and hashCode() methods for DDS member info in order to better differentiate the external Cassandra nodes.
SR-D22686 · Issue 493519
IH summaries working with external Cassandra
Resolved in Pega Version 8.4
Summaries were not working for external Cassandra. This was an issue with the IH summary component using the aggregated dataset as a reference, and this fix contains several components to improve this function. An issue where the IH aggregates dataset does not materialize when DDS is external Cassandra has been resolved by modifying the code that checks the DDS availability. A filter that was used to match “” did not work when pre-aggregation is off : this was due to the IH Browse operation being done outside of a DF context, and has been fixed. The IH Summary shape not working properly in a strategy if it referenced an aggregate dataset with pxInteractionID as part of the group-by properties was traced to an issue with it generating a pxInteractionID value when executing the strategy, and has been resolved by excluding pxInteractionID from the group keys in the IH Summary shape.
SR-D24880 · Issue 494175
CEP bucket size increased to 1 minute
Resolved in Pega Version 8.4
Frequent DSM Data Flow Errors were observed in production nodes with very high event rates. This was traced to the default bucket size for CEP being 1 second for a 31 day window, which was not sufficient to handle high volume use which may create an event strategy window large enough to hit the Cassandra column size limit. In order to resolve this, the default bucket size has been increased from 1 second to 1 minute.
SR-D26010 · Issue 500166
Modified VBD insertion logic to improve handling
Resolved in Pega Version 8.4
An issue with not being able to launch an additional VDB node was traced to two processes inserting into VBD with different field signatures and triggering unnecessary object creation. This was amplified when adding a second node as the objects were serialized. To resolve this, the insertion logic has been modified to avoid creating new data container/field descriptors after a new field is included, as well as a measurement with smaller data type than previous container.
SR-D26976 · Issue 507218
Filter added to ensure correct context for proposition strategy rules
Resolved in Pega Version 8.4
Given two applications (ex App1 and App2) hosted on the same domain where App2 was built on App1, trying to create a strategy rule in App1 and do a test run strategy using the propositional data component which internally uses App2 propositions generated the error: Failed to find a 'RULE-DECISION-DECISIONPARAMETERS' with the name 'GROUP_2'. There were 1 rules with this name in the rulebase, but none matched this request." Investigation showed the strategy was using the PropositionNoCacheUtils and PropositionTools java classes to load the propositions during run time. In these classes, the group classes were browsed from the db irrespective of the application context, causing the strategy run to fail as it was not able to access the decision data rules in other applications which shared the same SR class as the current application. To resolve this, a filter has been added to the PropositionNoCacheUtils and PropositionTools java classes to filter out the groups that are not in the current application context.
SR-D28038 · Issue 501757
External Cassandra nodes listed in DDS cluster
Resolved in Pega Version 8.4
Nodes of an external Cassandra cluster were not listed in the DDS cluster except for the first one in the host list, and when the “only” listed Cassandra node was restarted, the status on the DDS cluster LP did not become “NORMAL” afterwards. In addition, even though other C* nodes were up and running, the external Cassandra cluster was reported as unreachable. This was an unintended side effect of work done on the landing page to reflect the real state of the nodes after some were killed and restarted, and has been corrected by refining the equals() and hashCode() methods for DDS member info in order to better differentiate the external Cassandra nodes.