14.2 Job and Transaction Recording: Troubleshooting Faulted Unattended Jobs
Key Takeaways
Recording is a process setting on the Additional settings tab, disabled by default, and applies only to unattended automations.
Video captures the last 3 minutes, with options for failed jobs only, failed queue transactions, or all jobs.
Queue transaction recording requires the queue activities; status changes through HTTP or external APIs are not recorded.
Videos are kept 7 days (failed) and 3 days (successful) on Community, 14 and 7 days on Enterprise; screenshots are kept 30 days.
The Execution Media permission (View, Create, Delete) controls recordings, and recording activity is audited.
14.2 Job and Transaction Recording: Troubleshooting Faulted Unattended Jobs
Core Concept: Orchestrator can record the last moments of unattended executions as video or screenshots, so developers can see what happened on the robot's screen when a job or queue transaction failed. The exam description lists "Unattended automations features: recording".
Where to Enable Recording
Recording is configured per process, when you create it or later from the process settings. The Job Recording option is on the Additional settings tab and is disabled by default. The feature applies only to unattended automations.
Video Recording
Video captures the last 3 minutes of the execution. When enabled, choose one of these settings:
| Setting | What is recorded | Availability |
|---|---|---|
| Record and store only failed jobs | Only executions that fail | Automation Cloud Robots - Serverless, or Robot 2023.2 or later |
| Record and store failed queue transactions | Executions with failed queue transactions; the recording is shown with the transaction | Serverless, or Robot 2023.10 or later with System Activities 2023.10 or later; the process must use queues |
| Record all jobs | Every execution | Serverless, or Robot 2023.2 or later |
UiPath recommends turning on Record all jobs only briefly, to observe odd behavior, and then switching to failed-only recording to save bandwidth and storage.
Details that matter
- Queue transaction recording works only when the item status is changed with the dedicated queue activities, such as Set Transaction Status. Status changes made through HTTP activities or external API calls are not recorded.
- When a job is suspended and later resumed, the video includes only the execution after the resume.
Viewing a recording
- From the job's details, using Open recording, or the job's context menu.
- On the Video timeline tab, where logs are synchronized with the video: select a log entry to jump to that moment, filter logs by severity, and download the logs and the video.
- For a failed queue transaction, from the transaction's context menu or its details window.
How long videos are kept
| Plan | Failed jobs and failed queue transactions | Successful jobs |
|---|---|---|
| Community | 7 days | 3 days |
| Enterprise | 14 days | 7 days |
After that, videos are no longer available in Orchestrator; download any video you need to keep.
Screenshot Recording
Screenshots of the last moments before a failure are the lighter alternative:
| Setting | Meaning | Limits and default |
|---|---|---|
| Scaling | Screenshot scale in percent | Maximum 100; default 100 |
| Frequency | Interval between screenshots, in milliseconds | Minimum 250; default 500 |
| Duration | Seconds before the failure to capture | Maximum 120; default 40 |
Screenshots are retained for 30 days. The screenshot option does not work with Robots older than 2019.3 or with Cloud Robots - Serverless.
Permissions: Execution Media
Recording is controlled by the folder-level Execution Media permission:
| Permission | Allows |
|---|---|
| View | Watching and downloading recordings, including the Download option on the Jobs page |
| Create | The account that runs the job can create the recording |
| Delete | Deleting recordings |
The default folder roles that include Create on Execution Media are Folder Administrator, Automation User, and Robot. Video recording activity is included in the Orchestrator audit.
Recording Versus Other Diagnostics
| Tool | Best for | Limitation |
|---|---|---|
| Video recording | Seeing the UI state that led to a failure | Last 3 minutes only; limited retention |
| Screenshots | A light record of the last seconds | Up to 120 seconds before failure |
| REFramework exception screenshot | A single image at the moment of a System Exception | Only what the template captures, saved on the robot machine |
| Logs | The sequence of activities, values, and errors | No visual context |
| Healing Agent | Recommendations or self-healing for UI targeting failures | Covers UI automation issues only |
Recordings complement logs; they do not replace clear Log Message entries and meaningful exception messages.
Before You Enable Recording
- Confirm the robot and System Activities versions support the option you want.
- For queue transaction recording, check that the process sets statuses with Set Transaction Status rather than HTTP calls.
- Decide who needs Execution Media View and grant it only to them.
- Plan to download important videos before the retention period ends.
Privacy and Governance
Recordings may capture customer data shown on screen. Before enabling Record all jobs in production:
- Check whether the screens contain personal or financial data that policies restrict.
- Limit Execution Media View to the people who troubleshoot the process.
- Prefer failed-only recording in production.
- Remember that downloaded videos leave Orchestrator's retention control.
Scenario
A performer fails about once a week with "element not found" on a web portal, and the logs do not explain why. The developer enables Record and store failed queue transactions on the process. The next failure's video shows a cookie consent banner covering the button. The developer adds a Check App State step that dismisses the banner, and switches recording back off once the fix is verified.
Where is job recording turned on, and what is its default?
In Studio project settings; enabled by default.
On the machine template; enabled for all processes.
In the robot configuration file on each machine.
On the Additional settings tab of the process in Orchestrator; disabled by default.
A process records failed queue transactions, but a failure caused by a status change made through an HTTP Request to the Orchestrator API has no recording. Why?
Queue transaction recording works only when statuses are set with the dedicated queue activities, not through HTTP activities or external API calls.
Queue transactions can be recorded only on the Enterprise plan.
Recording starts only after the third failure.
Queue transaction videos are kept for only one hour.
Which permission lets a developer watch and download recordings of failed jobs?
Jobs Edit
Execution Media View
Logs Create
Monitoring Edit
Sections you finish are checked off in the contents.