ydb-platform
Database page isn't working during problems
Issue Description:
- The UI becomes stuck in an infinite loading loop when a database node repeatedly fails, rendering the page non-functional.
- The UI becomes stuck in an infinite loading loop when a Hive/SchemeShard repeatedly fails.
- As a result, users cannot view critical database information (e.g., name, metrics, logs) or the node list, ultimately preventing effective troubleshooting or administration.
Impact:
- Administrators are unable to access vital database details when issues occur, delaying diagnosis and resolution.
- The inability to view node statuses or leverage logs/metrics means any isolated node failure can escalate or remain unresolved for longer.
Expected Outcome:
- The database UI should always remain responsive, even if one (and only one) or more nodes or a system tablet are failing.
- Essential information––including database name, links to metrics/logs, system tablets list, and node uptime status––must be visible regardless of node health.
Proposed Fix: • Implement robust error handling or graceful degradation so that a single failing node does not obstruct the entire UI. • Provide clear indicators or alerts for any nodes that are failing, while still allowing access to healthy nodes’ information. • Maintain direct links to metrics and logs to facilitate swift debugging, even in degraded states.
Next Steps:
Update the UI logic to detect and isolate failing nodes, preventing them from causing infinite loading loops. Test thoroughly to confirm the UI remains accessible under various node failure scenarios. Monitor new logs and metrics post-deployment to ensure the fix performs under real-world conditions.
List of examples:
- https://st.yandex-team.ru/CLOUDDUTY-37887#68233541ab65a02dd6a47f81