If the Dgraph standard out log contains
connection broken
messages, although it may look like the
problem occurred with the Dgraph, the actual cause of the problem is usually a
broken connection between the server that hosts the front-end application and
the server that hosts the Dgraph.
In the case of connection errors, various parts of the Guided Search implementation issue the following error and warning messages:
The .NET API throws the following exception:
Endeca.Navigation.ENEConnectionException: Error reading from the connection. The operation has timed out at Endeca.Navigation.OptiBackendRequest.GetContent() at Endeca.Navigation.OptiBackend.GetNavigation(OptiBackendRequest req) at Endeca.Navigation.HttpENEConnection.Query(ENEQuery neq)
The Java API throws a similar exception.
The Dgraph standard out log contains warnings similar to the following:
WARN
[DATE TIME]
UTC (1239830549803) DGRAPH {dgraph}: Aborting request: connection broken: client 10.10.21.21And finally, the Dgraph request log contains an abnormal
status 0
message similar to the following:1239830549803 10.6.35.35 - 349 0 19.35 0.00 0 - 0 0 - -
Typically, the
connection broken
message means that the Dgraph
encountered an unexpected failure in the connection between the client and the
Dgraph. This type of error may occur outside the Dgraph, such as in the
network, or be caused by the timeout of the client application session.
Investigate the connection between the client and the Dgraph. For example, to prevent timeouts of the client application sessions, you may decide to implement front-end application retries.