Read or Write won't recover when write/read/lookup request is sent to coordinator server.

[loserwang1024]

Search before asking

• I searched in the issues and found nothing similar.

Fluss version

0.8.0 (latest release)

Please describe the bug 🐞

Currently, if we upgrade (or restart) multiple servers simultaneously, their pods may exchange IP addresses. This can cause issues because write/read requests might be sent to the wrong server.

If only multiple tablet servers are restarted, it is generally acceptable. In this case, when a server receives an incorrect write/read request and cannot find the leader, it returns an InvalidMetadataException. The client then updates its metadata accordingly.

However, if both the coordinator and tablet servers are restarted, the client throws an UnsupportedVersionException, which is not an InvalidMetadataException, so the client does not refresh its metadata. Unlike a NetworkException—which can be recovered from by retrying—the job will become stuck and fail indefinitely.

2026-02-04 20:12:28,595 ERROR org.apache.fluss.client.table.scanner.log.LogFetcher         [] - Failed to fetch log from node 49
org.apache.fluss.exception.UnsupportedVersionException: The server does not support FETCH_LOG(1015)

The server does not support LOOKUP(1017);

Solution

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!