This is a Major fix release. We recommend you update to this version ASAP. A bug might occur when a client performing a scan hits a "Partition Unavailable" during an unstable cluster (in both high availability (AP) and strong consistency (CP) modes). Previous versions of the client aborted the scan and put the connection back into the pool, which might cause unprocessed results to be sent to a different transaction (of the same client), possibly resulting in incorrect application behavior. This has been fixed by Go Client v5.10.0 and v6.4.0.
-
Fixes
- [CLIENT-1822] Scan/Query/Other streaming commands, including some Batch could put a faulty connection back to the pool after a cluster event where in certain conditions its buffer contents would end up in another scan and mix the results.
- Update go.mod, redact buggy versions and update required Go version to v1.16
- Fixes a few test errors for server v6.1+.
- Adds a
nil
check for error in the execution retry.
-
Improvements
- Update the examples for the new scan/queries.
- Avoid indirection for
[]byte
conversion during reflection. Resolves #382.