Description
Problem:
This is a branch of MB-40337 where the libcouchbase instance relies on the application to drive it's event loop, leading to issues like CBSE-16556 where the cluster topology can completely change between 2 lcb event loop progressions.
Patch in MB-40337 addresses the issue if the outgoing KV nodes (that are now no longer part of the cluster) is still hosting Couchbase service (the memcached process is still running) on the node, which triggers a LCB_AUTH_ERR which is used as a trigger to repair the lcb instances.
However, it's possible that the nodes are no longer running / unreachable from the eventing-consumer process.
Proposed Solutions:
- Given that this is can not be considered as a corner case scenario, either the libcouchbase event loop should be run on a separate thread.
- Or should consider ping operations during ideal time.