Queues and execution machines#

Platform-owned queue#

Queues and Tasks are Sherpa Orchestrator entities. Platform server logic selects, reserves, changes status and retries Tasks. Data is stored in the database. The built-in queue does not require a separate RabbitMQ service.

Selection and retries#

A Robot selects Tasks by identifier, status, priority, tags, locks and creation/update date ranges. Conditions can be combined in one request. Queue settings define automatic retry rules and the maximum number of attempts. Each attempt is reflected in processing status and logs.

Locks and permissions#

Tasks have separate read and write locks. A lock has an owner and expiry time and is released automatically when it expires. Acquiring a Task for processing protects against assigning it to several Robots simultaneously.

The Role, Account and access folder determine queue access. Viewing, editing and processing permissions are checked separately. One organization’s permissions do not expose another organization’s data.

Registration and execution mode#

A machine joins the organization through managed enrollment approval. Orchestrator creates the executor record automatically. Reconnecting uses the same identity without duplicates.

The platform determines a workflow’s GUI requirements from its blocks and subprocesses. Application interaction blocks require a compatible graphical workstation. Background workflows do not require an interactive desktop. Custom component execution requirements are checked with project dependencies.

Jobs, status and messages are available on the Jobs screen and in designer runs.