terrymanu commented on issue #35693:
URL:
https://github.com/apache/shardingsphere/issues/35693#issuecomment-3585495588
## Problem Understanding
- On 5.5.2 ShardingSphere-Proxy, with 100+ tenant databases, if any
Storage Unit datasource is unreachable during startup, the Proxy process exits,
blocking autoscaling. Repro: set frontend protocol to PostgreSQL, add an
unreachable PostgreSQL storage unit, start Proxy
→ startup fails.
## Root Cause
- During startup, Proxy initializes each Storage Unit by opening a
connection to read database product name/version (via DatabaseServerInfo); any
connection failure throws DatabaseServerLoadingException and aborts startup.
This fail-fast behavior is by design to ensure
configured datasources are reachable and metadata is available.
ShardingSphere focuses on traffic distribution/governance and does not perform
database liveness tolerance or skipping.
## Problem Analysis
- In multi-tenant setups with offline/not-ready databases, the startup
flow still attempts to connect every Storage Unit; the first failed connection
stops the process. Current versions do not support “ignore unreachable
datasource and continue” or lazy metadata
loading. Operators must ensure all Storage Units are reachable at
startup, or remove/disable unreachable ones before start and add them back
dynamically after recovery.
## Conclusion
- This aligns with current design and is not a ShardingSphere bug. Please
keep all Storage Units reachable before starting Proxy, or exclude offline
databases from the config and add them later when they are available.
ShardingSphere does not handle database state
awareness; availability and monitoring need to be managed externally.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
