> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/manage-and-monitor.md).

# Manage and Monitor

Monitor Cloud Streaming jobs, restart or stop processing, and configure connector offline notifications.

Use Cloud Streaming to check source health and Workflow state separately. An online camera does not mean a Workflow is processing it, and a Workflow waiting for video can remain enabled while its camera is offline.

## Web App

Open "Deployments", then "Cloud Streaming". Open a connector to see its sources, or a source to see its Workflow runs. A draft preview is listed apart from its deployment, so select the intended run before checking output, errors, or controls.

### Processing states

<table data-search="false"><thead><tr><th>State</th><th>What to check</th></tr></thead><tbody><tr><td>Waiting for video</td><td>No usable input. Check the connector or publisher.</td></tr><tr><td>Waiting for capacity</td><td>Waiting for a GPU. Extra jobs do not help.</td></tr><tr><td>Starting</td><td>Loading the Workflow and models.</td></tr><tr><td>Processing</td><td>Check that frame counts advance.</td></tr><tr><td>Reconnecting or Retrying</td><td>Recovering. Trackers and counters reset.</td></tr><tr><td>Adjusting resources</td><td>Restarting on a larger GPU after running out of memory.</td></tr><tr><td>Needs attention</td><td>Recovery paused after 5 failures, 15 minutes without processing, or 10 starts in an hour. Fix the cause, then restart.</td></tr><tr><td>Recovery suspended or Failed</td><td>Read the error before restarting.</td></tr><tr><td>Stopped</td><td>Does not resume on its own.</td></tr></tbody></table>

Runs without recovery (ex: previews) show raw states (ex: "Queued", "Running").

"Needs Attention" lists offline connectors and blocked Workflows. See [Troubleshooting](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/troubleshooting.md).

### Activity

Select a run and open "Activity" to see its latest lifecycle changes (up to 30 entries) when automatic recovery is on. It explains waits, retries, and failures. It is not prediction history, a video archive, or a log; use [Latest Result, sinks, or saved captures](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/consume-results.md) for those.

### Logs

Click "Logs" in the source's menu to see the processor output for its Workflow runs, including output from your Custom Python blocks. Each restart starts a new run. Logs use your Workspace's log retention and are billed like [device logs](https://docs.roboflow.com/deployment/self-hosted/enterprise/deployment-manager/monitoring/view-device-logs.md). While a run is active, you can also [download its recent results](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/consume-results.md#web-app) from this dialog.

### Restart processing

A restart applies the latest published Workflow and resets in-memory state, with exceptions described in [restart behavior](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md#restarts).

Select a run and click "Restart". The source list's restart skips previews. For an uploaded file, click "Run Again". If a run ran out of memory on the largest GPU, click "Choose a Workflow" and pick a lighter one.

To restart every active deployment, click "Restart All Streams", check the source and Workflow counts, and confirm "Restart All". Stopped streams stay stopped. If only some restarts succeed, "Try Again" retries just the unfinished ones.

Do not start new jobs while a restart is stopping the old run.

### Stop processing

Select the run, click "Stop", then "Stop processing". Stop each run you no longer need. Closing the page does not stop a deployment; see [stopping](https://docs.roboflow.com/deployment/roboflow-cloud/cloud-streaming/run-a-workflow.md#stopping).

### Rename or remove a source

Rename a connector-backed source from its action menu, or a direct source from "Connection Settings". Use "Delete" only to remove the source itself, not to stop a run. A connector that still reports a removed camera registers it again, so disable or remove the camera in the connector's configuration to keep it gone.

## Offline notifications

Notifications are opt-in at the Workspace level. Registering a connector alone does not send email.

1. In "Cloud Streaming", click "Offline notifications".
2. Enable notifications and select recipients from Workspace members.
3. Set the offline threshold, reminder choice, and hourly and daily limits.
4. Click "Save". Opening the dialog without saving does not enable delivery.

For a new configuration, the dialog selects the configuring member and defaults to a five-minute offline threshold, reminders enabled, two emails per hour, and ten per day. Existing disabled configurations stay disabled when opened. Limits apply to the Workspace, not independently to each connector.

Each email groups eligible offline streams across the Workspace. A new failure can trigger an update after its threshold, subject to the same limits. Reminders use increasing intervals: 30 minutes, then 1, 2, and 4 hours, then every 6 hours. A newly offline stream resets that schedule. No recovery email is sent.

Open a connector's page and use its "Offline notifications" toggle to exclude or include that connector. Connectors are included by default, but the Workspace must have saved, enabled notification settings.

### Notification coverage

Emails cover cameras on included connectors that run a deployment with recovery. Previews, direct sources, and blocked Workflows are not covered.

Emails can arrive later than the threshold.

Use "Send test email to me" to test. Acceptance does not confirm inbox delivery.

## Permissions

"View Stream Processing" allows access to source and run information. "Manage Source Video" is required to change source settings, start or stop processing, change notifications, or send a test email. Live preview and watch operations also require management authority because they request video publication. Saving a capture additionally requires upload permission.

Use [custom roles](https://docs.roboflow.com/platform/enterprise-features/role-based-access-control) to grant the access each operator needs. API integrations use the corresponding [API-key scopes](https://docs.roboflow.com/reference/authentication/authentication/scoped-api-keys).

## HTTP API

Read a job to check its current state and error. Do not infer success from a request timeout; inspect the existing job before retrying a state-changing operation.

{% openapi src="/files/HZKHykExD3Pp9vVK5PrS" path="/{workspace}/video-jobs/v1/{jobId}" method="get" %}
[cloud-streaming.yaml](https://1583372177-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FoksNmwps1HmYr8TqHDnX%2Fuploads%2Fgit-blob-2927c06848d4341631bbd3bafea073b855e32b47%2Fcloud-streaming.yaml?alt=media)
{% endopenapi %}

Cancel the job when processing should end, then read it until it reaches a terminal state. An active processor stops asynchronously.

{% openapi src="/files/HZKHykExD3Pp9vVK5PrS" path="/{workspace}/video-jobs/v1/{jobId}/cancel" method="post" %}
[cloud-streaming.yaml](https://1583372177-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FoksNmwps1HmYr8TqHDnX%2Fuploads%2Fgit-blob-2927c06848d4341631bbd3bafea073b855e32b47%2Fcloud-streaming.yaml?alt=media)
{% endopenapi %}

Restart, promotion, connector management, and offline notification settings have no API endpoints: use the Web App for those.
