Requests to gpfdist not spread across cluster segments

[Anton Belkin]

Hi, all!

We found that in our Greenplum cluster all requests to gpfdist, instead of being spread across cluster segments, are executed only on the gpseg-0 segment. External table has approximately the following content:

CREATE EXTERNAL TABLE stage.gpfdist_test_table (a text, b int)
LOCATION ('gpfdist://ip-10-100-100-100.eu-west-1.compute.internal:7000/demo/demo.gz')
FORMAT 'CSV' (delimiter ';');

All requests to above table has been executed only on the host where gpseg-0 is located and not on any other segments, parameter gp_external_max_segs  is set to 64 by default, in the cluster there are 218 segments. Greenplum version is 6.26.3

Does anybody faced similar issues before ?

This post was removed

[alex wood]

If requests to gpfdist are consistently reaching only one Greenplum segment instead of being distributed across the cluster, I would first check the external table definition, gpfdist host configuration, and the number and placement of gpfdist processes. It is also worth reviewing the segment-to-host mapping, network connectivity, and whether the external table is using a URI configuration that points all segments toward the same endpoint. Monitoring segment activity and gpfdist logs during a query can help determine whether the issue is caused by configuration or workload distribution. Keeping the cluster and gpfdist configuration consistent is important for balanced data loading and better performance.