Version upgrade - Dataphin - Alibaba Cloud Documentation Center

Upgrade Notes

Dataphin supports downtime, semi-downtime, and zero-downtime upgrades. Downtime upgrades use the legacy process, while semi-downtime upgrades use the new process. The following table describes the differences between these methods. For detailed instructions, see Downtime upgrade for Dataphin and Zero-downtime upgrade for Dataphin.

Differences	Semi-downtime upgrade	Downtime upgrade	Zero-downtime upgrade
Main upgrade steps	Select a version and upgrade method. Back up the database, perform a pre-upgrade check, upgrade the application, and update data. Verify the upgrade result.	Select a version and upgrade method. Stop submitting new nodes, force stop running nodes, stop the service, back up the database, upgrade the application, rerun the force-stopped nodes, and update data. Verify the upgrade result.	Select a version and upgrade method. Upgrade the application (automatic rolling upgrade).
Dataphin logon availability	Yes. Logon is available after the application upgrade is complete.	No	Yes

Impact of version upgrades

Downtime, semi-downtime, and zero-downtime upgrades have different impacts. The impact of each method is described below.

Semi-downtime upgrade

Logging in to Dataphin is unavailable during the upgrade. Running offline nodes, real-time nodes, and DataService Studio API calls are not affected. The submission of new nodes is prevented only during the application upgrade phase.

Offline tasks	Impact of semi-downtime upgrade
Running task instances	They run as normal. No impact.
Pending task instances	They can be submitted normally outside the application upgrade phase. They cannot be submitted during the application upgrade phase.
Instance monitoring for offline tasks	Alerting does not work during the upgrade. After the upgrade, alerting resumes, and alerts for events that occurred during the downtime are sent. If a run timeout alert rule is configured, an alert may still be triggered for a successful node. This happens because the node status is not tracked during the upgrade, which increases the total runtime.
Offline task management (modifying resource groups, starting and stopping tasks, and viewing logs)	Not supported because logon to Dataphin is unavailable.

Downtime upgrade

Logging in to Dataphin is unavailable during the upgrade. Running offline nodes must be stopped, and new nodes cannot be submitted. Running real-time nodes and DataService Studio API calls are not affected. The specific impacts are described below.

Feature	Feature point	Impact of downtime upgrade
DataService Studio	DataService Studio API calls	APIs for direct-connect data sources, service units, registered APIs, and composite APIs that do not reference logical table APIs can be called normally. APIs for logical tables, Dataphin data sources, and composite APIs that reference logical table APIs cannot be called during the downtime. During the application upgrade phase of the Dataphin upgrade (the DataService Studio application upgrade phase), a small number of requests may fail. A small number of requests may also fail during a zero-downtime upgrade. Only API calls in the production environment are supported. API calls in the staging (development) environment are not supported.
	Call authentication and permission verification	The system uses the cache for API call authentication and permission verification. Application authentication remains valid. This means that authentication for the application's AccessKey and AccessKey secret is effective during the downtime. If the caller provides an incorrect AccessKey or AccessKey secret, the authentication fails. API-level permission verification remains valid. However, API field-level permission verification expires after five minutes by default. This means that if an application has requested permissions for specific API fields, all fields become accessible after the field-level permission verification expires. If the system upgrade spans across midnight and an API permission expires during the downtime, the permission does not expire as scheduled. It expires only after the Dataphin application system is restored.
	API development	Modifying, publishing, and viewing APIs are not supported because logon is unavailable.
	API O&M	Viewing API operational status is not supported because logon is unavailable.
	API alerting	Throttling and alerting for APIs are not available.
Real-time jobs include real-time integration and real-time development.	Real-time execution of task instances	They run as normal. No impact.
	Real-time monitoring of task instances	Alerting does not work from the start of the downtime until the application is restored. After the application is successfully upgraded, alerting resumes. You do not need to wait for the Dataphin console to become accessible. Alert events that occur during the downtime are ignored.
	O&M for real-time tasks (such as modifying resource groups, starting and stopping tasks, and viewing logs)	Real-time node O&M is not supported because logon to Dataphin is unavailable.

Zero-downtime upgrade

A zero-downtime upgrade performs a rolling upgrade of the Dataphin application. During the upgrade, a small number of requests might fail. To resolve the issue, retry the request.