Details
-
Bug
-
Resolution: Fixed
-
Major
-
None
-
None
-
None
-
1
Description
If an agent is created with a single address which points to a non-kv node then we are failing to bootstrap against it. On the SelectBucket call we receive an ENO_ENT in response which causes the connection to fail itself and be retried. This means that CCCP then cannot use any connection to fetch the cluster map and so we cannot connect to the other nodes in the cluster.
In past versions, when we had bootstrap at the agent level, a failing select bucket would cause the agent to fall back to HTTP polling which would then manage to fetch the cluster map and be able to setup connections to the other nodes in the cluster.
We need a way to identify that the connection is failing and bubble that up to the CCCP poller so that it can fail itself and trigger the HTTP poller to run.