INC-212704 · Issue 707115
Explicit lock release added for Cassandra threads
Resolved in Pega Version 8.6.4
Customer Decision Hub was not responding for realtime request REST service calls, and thread dumps during start up were showing all HTTP request threads were stuck in CassandraSessionCache.getSession. If an error is thrown while invalidating an old Cassandra session, the system may fail to unlock the write lock. This results in subsequent threads being blocked on the session cache's ReadWriteLock when they attempt to retrieve the session from the session cache. To resolve this, an update has been made to ensure that invalidate session is wrapped in a finally block that releases the write lock and log any thrown errors.
INC-212754 · Issue 708688
Log4j file security vulnerability issue addressed
Resolved in Pega Version 8.6.4
A zero-day vulnerability was identified in the Apache Log4j logging software which could potentially allow malicious actors to take control of organizational networks. Pega has immediately and thoroughly addressed this issue. More information can be found at https://docs.pega.com/security-advisory/security-advisory-apache-log4j-zero-day-vulnerability .
INC-213176 · Issue 709974
Data flow stability improvements
Resolved in Pega Version 8.6.4
All nodes were showing as active and running in the cluster, including background, stream, web nodes, but the queue processor would not start. Log entries indicated "IllegalStateException: Invalid run transition found [PENDING_NODES->COMPLETED] on run [DF_OutboundPreProcessing_SMS] by [EventDescription{originator=CheckNodeAvailabilityTask, reason=Restoring previous state." Due to this, all dataflows were moved to queued state. As soon as the corrupted dataflow was deleted the other dataflows were resumed. To resolve this, data flow stability improvements have been added along with better handling of runs in Queued state.
INC-213189 · Issue 707344
Addressed performance issues for classless DSMClipboardPage
Resolved in Pega Version 8.6.4
The DSMClipboardPage implementation handles Page Group access as a classless page (i.e. pxObjClass = ""), creating potential performance issues in implementations where the page group is used with a very large number of distinct AccountIDs in a short period. This has been resolved by updating the system to not cache keys for page group classless page.
INC-213262 · Issue 709457
Refresh added to limit commit log partition size
Resolved in Pega Version 8.6.4
Issues were seen on DDS nodes, including slowness and randomly dropping out and being replaced by new nodes. WARN messages in Cassandra logs reported issues relating to a large partition size of adm_response_commit_log_date_tiered table. Investigation showed the large partition size (over 10GB) was causing frequent C* query timeouts due to scanning a large number of tombstones, resulting in node terminations. This has been resolved with an update which will limit the commit log partition size by refreshing the NID every set interval.
INC-213945 · Issue 709365
Additional logging and executor work for Adaptive Decision Manager
Resolved in Pega Version 8.6.4
The Adaptive Decision Manager Pulse operation stopped running on some of the ADM nodes, causing an impact on the model update. To address this, some additional protections have been added for pulse operations running in a scheduled executor, and ERROR logging has been enabled.
INC-216894 · Issue 712241
Refresh added to limit commit log partition size
Resolved in Pega Version 8.6.4
Issues were seen on DDS nodes, including slowness and randomly dropping out and being replaced by new nodes. WARN messages in Cassandra logs reported issues relating to a large partition size of adm_response_commit_log_date_tiered table. Investigation showed the large partition size (over 10GB) was causing frequent C* query timeouts due to scanning a large number of tombstones, resulting in node terminations. This has been resolved with an update which will limit the commit log partition size by refreshing the NID every set interval.
INC-183960 · Issue 675870
Condition check added to trigger search update
Resolved in Pega Version 8.6.4
Following work done to update deleting attachments for cases created through an email bot, search indexing was not updating automatically and a manual re-index was needed to allow the global search to work. This has been resolved with the addition of a condition check which will update the case when there is a delete operation done on the link page.
INC-198725 · Issue 707482
Elasticsearch will split queries to avoid URL length limits
Resolved in Pega Version 8.6.4
Elastic Search Functionality was not working and showed the status as Offline. Investigation showed that when an external Elasticsearch setup had an index prefix length = 200 and a large number of dedicated indexes, the error "too_long_frame_exception, reason: An HTTP line is larger than 4096 bytes" was generated. In this case, the system was working with approximately 87 dedicated indexes and the name for each of them was sent over the URL, resulting in this error message. This has been resolved by updating the system to split queries so as to not exceed the HTTP line limit in Elasticsearch.
INC-200973 · Issue 699266
Added handling for client-side validation errors in sub-reports
Resolved in Pega Version 8.6.4
When a sub-report was trying to access externally mapped fields or an optimized page property by JOIN of another class, the main report was unable to validate the Page property. This has been resolved by updating the activity code to handle client-side validation errors on main reports attempting to access the externally-mapped Page properties of another class in the sub-report.