INC-152057 · Issue 621211
S3 attachment migration handles LInk-Attachment with multiple instances
Resolved in Pega Version 8.6
After S3 attachment migration, some attachments were intermittently not opening and displaying an error relating to being unable to load the file. Investigation showed that the attachments that failed to open did not have a pxStorageType tag in the XML of the work item. After the migration is done, the corresponding Data-WorkAttach-File and Link-Attachment instances are updated to point them to the repository. In this case, multiple LInk-Attachment instances were pointing to the same Data-WorkAttach-File instances, so only one Link-Attachment was updated and all of the other instances pointing to the same Data-WorkAttach-File instance were rendered unusable. While there was a workaround of manually updating the storage type in the database, this has been resolved by updating "pzgetattachmentcontent","pzupdatesourcereferences", and "pzuploadcontenttoexternalstorage" to check whether attachment is already migrated or not using the boolean parameter "isAttachmentAlreadyMigrated" to ensure all of the other Link-Attachment instances for a particular attachment are updated. If a migration was done and all Data-WorkAttach-File instances are pointing to a repository with some Link-Attachment instances not updated, those will be updated by running the migration again.
INC-152170 · Issue 609058
Resolved hung thread potential
Resolved in Pega Version 8.6
Attempting to create an empty file in the S3 bucket using the S3 repository create file API was causing threads to hang in RUNNING state forever, leading to the node failing a health check and generating warnings of "Detected stale thread for node <batch node id>". This has been resolved by updating the Java concurrency classes.
INC-152435 · Issue 612995
Hardcoded cluster password deprecated
Resolved in Pega Version 8.6
An unneeded hardcoded cluster group password has been removed in 3.8 version of Hazelcast as this use has been deprecated.
INC-152440 · Issue 614333
Compiler jars load as expected
Resolved in Pega Version 8.6
The system was not able to pick up the Default Paths and Default Classes arguments in Configure->System->Settings->Compiler tab when attempting to use a third-party custom jar file. This was an unintended consequence of work done to address performance issues when a requested DASS instance was missing, and has been resolved by ensuring that null values are not cached where are are lookup failures. In addition, enhanced logging has been added to SystemSettingsImpl.
INC-152634 · Issue 616757
FileListener correctly starts with debug on for FileS3Adapter
Resolved in Pega Version 8.6
When debug was enabled for class com.pega.platform.physicaldatalayer.internal.cloud.aws.s3.FileS3Adapter, no FileListeners could be started from admin console. The error "File.FoundationDataFileIngestion - Unexpected exception caught during processing" was generated. This has been resolved with the addition of a null check on publicAPI before starting the timer.
INC-152647 · Issue 609603
Email Listener auto-reply evaluation updated
Resolved in Pega Version 8.6
After upgrade, messages were being read but not getting processed for a specific Email listener (RCEmailListerner). The error "Email flagged as an autoreply email and will not be processed" appeared in the logs. Previously, an email was not considered to be an auto-reply only when the 'auto-submitted' header didn't exist or existed with value 'no'. This caused issues with auto-forward or auto-redirect emails where 'auto-submitted:auto-generated' could be in the header. Due to this, email was marked as auto-reply and email listener stopped processing it. To resolve this, the system has been modified to mark the message as auto-reply if it finds 'auto-submitted: auto-replied' in the header, but not 'auto-submitted:auto-generated'.
INC-153014 · Issue 625696
Handling added for missing archival class definitions
Resolved in Pega Version 8.6
Handling has been added to avoid suspending the archival process when a class definition no longer exists in the system. If the system does not find a class corresponding to a configured case type, the exception generated will be logged and processing will continue with the next case type.
INC-153086 · Issue 619322
Job scheduler tracing updated for multi-node
Resolved in Pega Version 8.6
When attempting to trace the job scheduler using "trace any node", the error "Unable to trace trace job scheduler 'ConsumerPathRequestEvents' on node as the scheduled tracer information was reset" appeared intermittently. When Job Scheduler is configured to run on "any associated node" and the trace is designated as "trace any node" option, the implementation was running the trace on the node that received the request which may or may not be the one running the job scheduler. This was not an issue when tracing a job scheduler that was running on "all associated nodes". To ensure the correct node is found, when using "trace any node" for the job scheduler, the system will now look for the eligible node and return the requestor ID for the trace.
INC-153310 · Issue 611881
Connectivity issue recovery made more robust
Resolved in Pega Version 8.6
In order to ensure agents are able to recover from a problem connecting to the database due to network issues or some other transient problem, the retry logic has been updated to classify connection-related exceptions as temporary issues and retry the agent's run.
INC-153664 · Issue 613053
QP general metrics now disabled by default for better performance
Resolved in Pega Version 8.6
While testing decisioning on fnx cloud platform, using an application to simulate load and using the Stream service with Kafka Instaclustr as a backend, large latency spikes over 1,000 ms were observed at around 33 second intervals from Stream writes. Investigation identified that the metrics that are gathered for every QP every 30 sec (on every single node) include a "lag" calculation that requires a heavy process operation during which more threads are spun off, causing CPU spikes. To mitigate this, the QP general metrics are now disabled by default and can be enabled using DSS.