Pega® Predictive Diagnostic Cloud provides insight into your system's health and performance; brings visibility to performance, stability, and quality issues; and simplifies root cause analysis. Predictive Diagnostic Cloud receives, analyzes, and stores data that is sent from Pega Platform. The information is gathered from several data sources and received as summary statistics or individual data.
Pega Platform uses a whitelist model for sending clipboard parameter data. This means that only the parameters that are required for analysis and that have known and safe content are sent. If a parameter is not listed as safe, the parameter name value is removed. The following parameters are on the whitelist by default: AJAXTrackID, ActivityClassToExecute, ActivityNameToExecute, CustomActivityClassName, CustomActivityName, FlowClass, FlowType, Format, InsKey, RuleClass, RuleObjClass, StreamClass, StreamName, TaskStatus, ViewClass, ViewInsKey, ViewOwner, ViewPurpose, action, actionName, activityName, contentID, currentLockOwner, dynamicContainerID, flowType, harnessName, inStandardsMode, insName, objClass, openHandle, originalLockOwner, portal, portalName, portalThreadName, preActivity, primaryPageClass, productName, productVersion, pxObjClass, pyAction, pyActivity, pyClassName, pyDefinitionKey, pyExecuteOnDataPage, pyForEachCount, pyPageName, pyReportClass, pyReportName, pyRuleset, pyRunType, pyStream, pyStreamName, pyTempPlaceHolder, pzTransactionID, requestorID, tabIndex.
Pega Platform sends information to Predictive Diagnostic Cloud from the following channels:
- Alerts and exceptions – Asynchronously sent to Predictive Diagnostic Cloud and written to the local alert files. Predictive Diagnostic Cloud receives most of the data that is written in the PegaRulesAlert log file. Only business sensitive data is filtered out and removed.
- PegaRULES ManagementDaemon thread – Sends a node health status message to Predictive Diagnostic Cloud every two minutes. The health status message contains current health statistics such as CPU utilization, memory utilization, agent count, requestor count, recent responsiveness, and the time when the last system pulse agent ran. No work data is included in the health status message.
- PegaAESRemote agents – Run on all monitored nodes and periodically send information to Predictive Diagnostic Cloud to assess the overall health and help determine any issues.
In Pega Platform 7.3.1 and earlier versions, the PegaAESRemote agents sent only the following information:
- Summarized usage data – The PushLogUsage agent runs a report against a log-usage class to assess user count, interaction count, and average response time.
- Schema data – Database table name and index definitions. This data assists with the analysis of slow database queries.
- Guardrail statistics – Rule warning count.
Beginning with Pega Platform 7.4, a series of PegaAESRemote agents gather the following data for Predictive Diagnostic Cloud:
- Database query statistics – An agent gathers, resets, and sends Postgres database query statement statistics. In Predictive Diagnostic Cloud, this information is available for administrators to assess database load that could be running in the background.
- Database table statistics – An agent gathers and sends Postgres table sizes and access statistics. In Predictive Diagnostic Cloud this information helps administrators understand and explore research usage, and improve analysis of slow database queries.
- Pega run-time environment – An agent gathers Java code and sends Pega configuration options to assist with solving problems through Predictive Diagnostic Cloud.
- Java virtual machine - An agent gathers the basic run-time information about the Java virtual machine that is used on the node.
Predictive Diagnostic Cloud collects the following data from monitored nodes:
- Alerts – Most of the data sent to Predictive Diagnostic Cloud comes from Pega alerts that are written to the PegaAlerts log file. Over 50 types of alerts are triggered when particular counts or elapsed times exceed a threshold during an interaction. Alerts contain metadata about what happened in an interaction. For more information about metadata that is included in alert logs, see Alert log message data. For more information about the types of alerts, see Performance alerts, security alerts, and Autonomic Event Services.
- Parameter page – Some parameters from the current parameter page are sent along with the alert. The parameter page contains important contextual information about the functions that run in the monitored application. The required parameters are sent to a whitelist in Pega Platform. All remaining parameters are filtered out and excluded. You can configure the whitelist to send additional parameters as necessary to provide more context for your alerts.
- Database alerts – The database query is sent as part of the alert. All business data values are removed in the SQL calls with INSERT INTO statements.
- Exceptions – Exceptions are sent to Predictive Diagnostic Cloud for analysis (DEBUG or INFO statements are not sent). You can view exceptions by looking at the error lines in the PegaRULES log file. Exceptions can contain some contextual data that is a subset of the fields sent for alerts.
- Performance statistics – The monitored application sends performance statistics hourly. Predictive Diagnostic Cloud uses these statistics to identify overall performance and performance trends of your application’s systems, including statistics for average response time and unique user count.
- Database indexes – Predictive Diagnostic Cloud gathers the database index information daily to create recommendations for improving query performance. You can see which indexes are currently used and determine whether you need additional indexes. The PushDBIndexes activity gathers the index information for all tables and sends it to Predictive Diagnostic Cloud.
- Guardrail violation counts – Guardrail warnings indicate that the rules in your application do not follow all Pega Platform best practices. Predictive Diagnostic Cloud counts the total number of rules and also counts individual rules that have justified and unjustified warnings of severe, moderate, or caution.
All communication with Predictive Diagnostic Cloud is fully encrypted because the data is transmitted through a SOAP protocol over HTTPS. Your application sends SOAP messages to Predictive Diagnostic Cloud; Predictive Diagnostic Cloud cannot reach back to your application for additional information. The multitenant features in Pega Platform ensure that the customer (tenant) data can be accessed only by that customer. This means that the information flow between your monitored application and Predictive Diagnostic Cloud is one way. Users can use Predictive Diagnostic Cloud by directly logging in to Predictive Diagnostic Cloud through a web browser over an encrypted HTTPS connection or by subscribing to emailed reports. For more information about multitenancy, see the Multitenancy Administration Guide.
Predictive Diagnostic Cloud is an application built on Pega Platform and securely hosted in a dedicated private cloud on Pega Cloud. Predictive Diagnostic Cloud uses multitenancy capabilities by giving each customer a unique URL, user name, and password. All customers are segregated and have access only to their system database on their unique URL and can never see information or data about other customers.
Data handled by Predictive Diagnostic Cloud is stored in persisted memory and encrypted with a 256-bit AES key. The keys are automatically rotated periodically, securely stored in an encrypted key management system (KMS), and managed by the Pega Cloud service.