> ## Documentation Index
> Fetch the complete documentation index at: https://docs.casebender.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Integration Operations

> Monitor connector health, action outcomes, catalog versions, and credential rotation.

## Health tab

Open **Settings → Integrations → Health** to review the integration execution
plane. The Health tab summarizes:

* Enabled and unhealthy connections
* Connector action success and failure counts
* Recent execution latency and errors
* Credential health and rotation state
* Installed catalog size and connector maturity
* Jira outbox and webhook status counts for the last 24 hours
* Jira policy version, sync health, and recent recoverable failures

Select **Manage** on a Connections row, then **Test connection**, to run the
connector's declared health action. Testing uses the same host allowlist,
timeout, response limit, tenant isolation, and redaction controls as production
execution.

## Jira recovery

Jira outbound automation is recorded in a durable tenant-scoped outbox.
**Retry now** makes an eligible failed event immediately available without
changing its remaining attempt budget. **Replay** moves a dead-letter event back
to pending and grants the displayed additional attempts. Both actions preserve
the event's original policy snapshot, payload, correlation context, and
idempotency key, and both write an immutable unified audit record.

Confirm the mapping or provider problem is resolved before recovery. Recovery
controls are available only to users with existing integration-management
permission, and event identifiers are resolved inside the active organization.

Jira webhook processing uses an expiring ownership lease. **Release lease** is
available only when a `PROCESSING` receipt is stale. Releasing it marks the
expired claim failed so the existing queue retry or Jira redelivery can claim
the receipt. It does not republish or reconstruct the webhook: raw inbound
payloads are intentionally not persisted.

Use Jira's delivery log to redeliver a failed webhook when the worker queue no
longer has the original message.

## Connector maturity

Catalog entries can be **certified**, **supported**, **community**, **private**,
or **deprecated**. Maturity is version-specific. Casebender pins the connector
version when creating a connection.

## Troubleshooting

1. Confirm both `workflow-processor` and `connector-worker` are healthy.
2. Confirm they use the same Redis action and result queue names.
3. Check that the connection is enabled and visible to the workflow's
   organization.
4. Verify that a live credential is bound and that its host allowlist covers the
   configured base URL.
5. Review the connector worker error and circuit-breaker state.
6. Rotate an expired credential, then retest the connection.
7. For Jira, inspect failed/dead-letter automation and failed, rejected, or
   stale webhook receipts. Resolve the cause before using recovery.

Connector responses are bounded and secret fields are redacted before outputs
are persisted. Use vendor-side request identifiers for deeper investigation;
do not enable raw secret or response-body logging.

For emergency rollback, disable the connection to stop new workflow and
automation use. Disable individual Jira policy controls when only one event
class must stop. Restoring previous mappings publishes a new policy version;
historical outbox events remain immutable.