Troubleshooting connection timeouts

To help prevent timeout issues during data ingest operations, use the --net-timeout flag for the data domain profile's configuration.

At least one Dgraph process of the Oracle Endeca Server starts for each data domain. It has a default request timeout of 30 seconds. This setting determines the maximum number of seconds that the Dgraph process waits for the client to download data from queries across the network. The client can be an end user sending a query to the Oracle Endeca Server, or for data ingest operations, an ETL client program that is loading records into the data domain.

If the client opens a connection with the server, the server will wait (for the length of the timeout period) for the receipt of client data on that socket. If the client does not send data within the timeout limit, then the server will drop the connection and log an HTTP 408 error in the Dgraph log.

For the data loading operations, this timeout limit may pose problems if you have a Data Ingest Web Service client that takes longer to send data. If the timeout limit is exceeded, the ingest request fails (because the server closes the connection), and the record batch is not loaded into the data domain.

If you continually see HTTP 408 errors in the logs, first verify that your ETL client is working properly. For example, make sure that the program is not spending an unusual amount of time in an operation that would cause it to exceed the timeout limit.

If you believe that the ETL client is executing as expected, but needs a longer request timeout period, then you can try increasing the request timeout setting. Use the --net-timeout flag to set the request timeout to a number that works for the ETL client. You will probably have to experiment with several settings to find the one that is optimal for your needs. The following command for the data domain profile sets this timeout: endeca-cmd --put-dd-profile --net-timeout <value>.