This topic describes the major feature changes and bug fixes released for Realtime Compute for Apache Flink on July 29, 2026.
This release is being rolled out in phases across the network. For upgrade progress, check the latest announcements in the right pane of the Realtime Compute console. If the described capabilities are not yet available in your account, the grayscale upgrade has not reached your account yet. To expedite the upgrade, submit a ticket with your business requirements, and we will schedule accordingly.
Overview
This release focuses on three areas: filling out enterprise-grade security, opening AI capabilities to third parties, and improving the operations experience. On the platform side, Connector credential-free access and a high-availability disaster-recovery drill are added, raising the security and compliance baseline. On the OpenAPI side, endpoints for the AI Assistant, autopilot policy management, skill queries, and data queries are opened, so large customers can integrate Flink AI and data capabilities into their own platforms. On the experience side, the job operations list gains multi-dimension filters and a restart-count column, and the health-score diagnostics adds a restart-count rule, shifting operations from reactive response to proactive prevention.
Platform capabilities
High-availability disaster-recovery drill
High-availability namespaces support initiating an AZ-level disaster-recovery drill. The system automatically evicts and relaunches job nodes, restores job status, and monitors the migration in real time. When the drill completes, a report is generated with core metrics such as RTO and job recovery rate. High-availability customers can now run drills without operator involvement to meet compliance audits such as MLPS or ISO, turning the process from manual operations into a one-click, visual, and auditable workflow on the console.
Credential-free Connector access (public preview)
Adds support for referencing RAM roles, integration with KMS credential management, and a refined variable management entry. The OSS Connector supports accessing data sources through a RAM role, and the PostgreSQL-CDC Connector supports accessing KMS credentials through a RAM role. Job configurations no longer need to carry AccessKey/SecretKey or database passwords, while existing explicit key configurations remain supported. Sensitive credentials are decoupled from job code, plaintext key risk is eliminated, security audit pressure is reduced, and key rotation no longer requires per-job changes.
Experience improvements
Job operations enhancements
The job operations list now supports filtering by engine version and job type (SQL / JAR / Python / YAML) and shows a restart-count column. These changes address the lack of filtering when job counts grow, the invisibility of internal restart anomalies, and the difficulty of locating batch anomalies by time, shifting operations from reactive response to proactive prevention.
Health-score diagnostics enhancement
Adds a restart-count health-score rule: cumulative restart count > 0 rated as Medium, > 100 as High, and > 300 as High. This fills the gap in scoring continuous metrics and eliminates the blind spot of "hundreds of restarts still scoring 100", making risks visible and scores consistent.
OpenAPI
AI Assistant OpenAPI
Opens the full lifecycle interface of the AI Assistant with an async-first design that supports submitting a session, querying status, and fetching results in asynchronous mode, along with an SSE streaming interface for real-time end-user conversations. One new operation is added:
-
Start a conversation (
ChatAiAgent)
Autopilot OpenAPI
Opens the Autopilot policy management and tuning-history query interfaces, allowing customers to configure tuning policies in bulk and track results from their own platforms. This closes the last gap in the automation loop and eliminates the need to switch to the console for manual operations. Three new operations are added:
-
Query Autopilot policy (
GetAutopilotPolicy) -
Update Autopilot policy (
UpdateAutopilotPolicy) -
List Autopilot tuning histories (
ListAutopilotTuningHistories)
Data query OpenAPI
Opens the full lifecycle management interface of SQL data-query scripts, so external systems can execute SQL queries, fetch results, and manage script files through API. Six new operations are added; combined with the existing StartSqlExecution operation, they form a complete closed loop of the SQL data-query API, enabling direct querying of Flink data from your own platform without logging in to the console:
-
Stop a query execution (
StopSqlExecution) -
Fetch query execution results (
FetchSqlExecutionResult) -
Create a query script (
CreateSqlFile) -
Get a query script (
GetSqlFile) -
Update a query script (
UpdateSqlFile) -
Delete a query script (
DeleteSqlFile)