Pega Robotics Orchestration - Key Concepts (Notes)

Vocabulary flashcards covering orchestration concepts, roles, patterns, rollout, observability, security, and best practices from the lecture notes.

Orchestration

The managed flow that coordinates when and where automations run across Pega Platform and Robot Manager; the control plane for unattended RPA and attended RDA, governing capacity, versions, and exceptions.

Robot Manager

Admin application for packages, versions, work groups/queues, robot inventories, routing, schedules, and dashboards used to govern orchestration.

Work group

A pool of robots eligible to work one or more work queues, used for capacity management and version pinning.

Work queue

Backlog of robotic assignments produced by cases (formerly workbaskets) awaiting processing.

Package

A deployed bundle from Pega Robotics Studio exposing named Entry Points and Attended Activities.

Entry Point (unattended)

The automation method a robot runs for a queued item in unattended processing.

Attended Activity (attended)

Desktop automation triggered from the user’s case UI during the user’s session.

Robot Runtime

The runtime engine that registers robots, manages sessions, and executes work.

RPA Service

Windows service that supports Robot Runtime by registering robots, managing sessions, and executing automation.

Case orchestration shapes

Shapes used to orchestrate cases: Queue for Robot (RPA) and Run Robotic Automation (RDA); names vary by release.

Dispatcher–Performer pattern

A layering approach for queues where dispatchers route work to performers, enabling organized work flow and DeadLetter handling.

Dead Letter queue

A separate queue for failed items that require triage or manual intervention.

SLA-driven prioritization

Prioritizing work based on service level agreements to meet critical deadlines.

Deployment rings

Staged rollout method: Pilot → Canary → Broad rollout via targeted work groups.

Version pinning

Locking a package version to specific work groups or robots to control upgrades.

Pilot

A small initial group used to test a new version before wider deployment.

Canary

A middle rollout stage to validate performance with a subset before full release.

Broad rollout

Full deployment to the intended population after successful pilot/canary checks.

Deployment rollback

Reverting to a previous package version by re-pinning, without rebuilding packages.

Version control

Managing package publishing and version pinning per work group or robot.

Capacity

Number of robots per work group and the management of their sessions.

Backpressure

Mechanisms to throttle throughput by queue separation and limited robots, with per-target concurrency control.

Token (concurrency token)

A concurrency control method where a token is acquired before performing work on a target to cap parallelism.

Retries

Configured retry attempts for failures within assignments/flows, often with backoff.

Exponential backoff

Increasing delay between retry attempts to reduce retry storms.

Poison detection

Identifying consistently failing items and moving them to quarantine for triage.

Observability

Dashboards, run histories, logs, and correlation IDs to monitor and troubleshoot orchestrations.

Correlation ID

A unique identifier that links case activity to robot logs across the workflow.

Runtime logging

Structured logs including Entry Point name, sanitized inputs, ReasonCode, and timing.

SLO dashboards

Dashboards that track Service Level Objectives such as success rate and latency.

Security

Controls for operator identity, RBAC, environment separation, and secure handling of credentials and PII.

RBAC

Role-Based Access Control; least-privilege access with separation of duties across environments.

Secrets vault

A secure store for credentials used by robots and workflows, outside package payloads.

PII redaction

Masking or omitting personally identifiable information in logs and payloads.

DR/HA

Disaster Recovery and High Availability planning, including backups, failover/runbooks, and tested restores.

RTO/RPO

Recovery Time Objective and Recovery Point Objective per process tier.

RPS

Robotics Package Server; legacy package hosting that may be migrated to Robot Manager.

HealthCheck Entry Point

A lightweight entry point used to verify package health on a canary robot.

Health checks

Lightweight tests that pause queues or adjust schedules when dependencies fail.

Exceptions taxonomy

Classification of errors into System, Business, and Data quality exceptions with defined actions.

System exceptions

Transient failures (timeouts, selector issues, session loss) retried with backoff and quarantined if needed.

Business exceptions

Deterministic failures (missing data, validation errors) not retried; routed to human with ReasonCode.

Data quality exceptions

Data issues that are corrected (enrichment/cleansing) and requeued after fix.

ReasonCode

A code describing why a case followed a particular exception path, used with messages.

Message

Human-readable explanation accompanying a ReasonCode for operators.

Attended vs unattended selection

Guidance: use unattended RPA for high-volume, back-office tasks with SLAs; use attended RDA for real-time user assistance.

Queue for Robot

A dedicated work queue used for unattended automation assignments via Entry Points.

Run Robotic Automation (RDA)

Attended automation triggered from the user’s desktop UI that returns results to the case.

Start My Day

An action to launch prerequisite desktop apps as part of a user’s session during RDA.

Entry Point inputs/outputs contract

Defined, small, stable input/output contract for an Entry Point to ensure reliability.

DataPage/JSONRef

Structured outputs (DataPage or JSON reference) returned to the case context.