To provide better services to users, Realtime Compute for Apache Flink builds a lifecycle management system in which product lifecycle policy (PLP) is defined to manage the service type changes and runtime lifecycle policy (RLP) is defined to manage the engine version changes. The lifecycle management system helps meet your increasing requirements of real-time business and ensures that your business runs in an efficient and stable manner by developing new services and versions.
Scope
PLP and RLP are applicable to all service types and versions of Realtime Compute for Apache Flink. For more information about the features and release status of each service type and version, see Service types.
Description of PLP
PLP is used during the process of a service type of Realtime Compute for Apache Flink from the official launch to delivery, to end of marketing (EOM), and finally to end of service (EOS). PLP is used for Realtime Compute for Apache Flink only when Realtime Compute for Apache Flink has a major service change. For example, PLP is used when Blink is updated to Realtime Compute for Apache Flink.
The period of time for each phase of PLP is not fixed. Before the Realtime Compute for Apache Flink team decides to conduct PLP, the team considers various factors, such as the maturity of a new service type, migration demands of users, smoothness of the migration process, and changes in costs of the new service type for users. If the team determines to conduct PLP, the team notifies users at least four months in advance. For more information, see the Related services and notifications section of this topic. The following figure shows the discontinuation process in PLP of Realtime Compute for Apache Flink.
Discontinuation process | Description | Remarks |
Announcement to EOM | Users are notified at least four months before the EOM phase by using a notification channel, such as announcements, internal messages, emails, or text messages. |
|
EOM to EOFS | EOM consists of the following phases:
Note The entire EOM phase cannot be less than four months. |
|
EOFS to EOS | New versions and patch versions are no longer released. Services are no longer provided, such as answering questions, resolving issues, and handling service level agreement (SLA) guarantee compensation. |
|
EOS | The service is discontinued and all support services are no longer provided. | This service type no longer provides services. |
Description of RLP
Realtime Compute for Apache Flink uses the enhanced engine Ververica Runtime (VVR). For more information about the engine version planning and definition, see Engine version.
RLP is used during the process of the VVR engine in Realtime Compute for Apache Flink from the official delivery to users, to EOM, and finally to EOS.
The period of time for each phase of RLP is not fixed. Before the Realtime Compute for Apache Flink team determines to conduct RLP, the team considers various factors, such as the compatibility and maturity of a new version and the demands of users for version updates. If the team determines to conduct RLP, the team notifies users at least three months in advance. For more information, see the Related services and notifications section of this topic. The following table describes the milestones in RLP of Realtime Compute for Apache Flink.
During the version iteration of the VVR engine, the final minor version of a major VVR version is the version of long-term support (LTS). For example, the final minor version of VVR 6.x is VVR 6.0.7. In this case, the LTS version of VVR 6.x is VVR 6.0.7. Only two LTS versions are maintained by Alibaba Cloud. If a new LTS version is released, the oldest LTS version no longer provides services. This indicates that you cannot select this VVR version when you create a new deployment of Realtime Compute for Apache Flink.
If a VVR version has critical security or stability issues, Alibaba Cloud shortens the RLP of the version to three months to avoid severe impact on business caused by this version, and notify the involved users by using the Realtime Compute for Apache Flink console or by sending announcements in a timely manner to update the version.
Milestone | Description | Remarks |
General availability (GA) | The new version can be officially released and delivered to users for use in the production environment. |
|
EOS | Realtime Compute for Apache Flink no longer provides the related services and support for this version. | You can no longer create deployments of this version in Realtime Compute for Apache Flink. The running deployments of this version are not affected. However, you are liable for any business losses caused by the defects of this version. Important If a major defect occurs in a minor version and the defect causes major losses to users, such as losses due to security issues or data accuracy, the Realtime Compute for Apache Flink team can unpublish the minor version and update the version to a new compatible version. |
Related services and notifications
Based on the preceding lifecycle policy rules, Alibaba Cloud performs the following operations:
Plan the lifecycle of each service type and engine version of Realtime Compute for Apache Flink and provide information to users during service consultation and the use of service.
Notify users at least four months before the specific time for each milestone by using a notification channel, such as announcements, internal messages, emails, or text messages. Notify users at least three months before the specific time for each milestone of the engine version.
Help users evaluate the risks of running deployments and provide solutions before the service type or engine version service is terminated.
References
For more information about the release notes for Apache Flink engine versions and platform upgrades, see Release notes.
For more information about the enterprise-level state backend storage GeminiStateBackend and the performance comparison between GeminiStateBackend and RocksDBStateBackend, see GeminiStateBackend.
For more information about the version updates or product activities of Realtime Compute for Apache Flink, see Service notices of Realtime Compute for Apache Flink. You can also log on to the Realtime Compute for Apache Flink console to view the version updates or product activities.