"interconnect encountered a network error" Displayed When Greenplum Is Running SQL Statements

Symptom

Before KAEzip is enabled, the messages "interconnect encountered a network error, please check your network (xxx)" and "DETAIL: Failed to send packet (xxx) after xxx retries in xxx seconds" are displayed when Greenplum is running SQL statements.

Key Process and Cause Analysis

The general cause is the loss of network data packets.

Conclusion and Solution

1. While ensuring that the hardware network is normal, significantly increase the values of the parameters ipfrag_time, ipfrag_high_thresh, ipfrag_low_thresh, and ipfrag_max_dist. The reference configurations are as follows:

   net.ipv4.ipfrag_max_dist = 64000
   net.ipv4.ipfrag_time = 6000
   net.ipv4.ipfrag_high_thresh = 6710886400
   net.ipv4.ipfrag_low_thresh = 671088640

2. Run the SQL statement again.
3. If the fault persists, set the value of gp_max_packet_size of Greenplum to a value smaller than the default MTU value of the server NIC.

   In this example, the default MTU value is 1500. Therefore, set the value of gp_max_packet_size to 1492:

   gpconfig -c gp_max_packet_size -v 1492

4. Optional: Run the SQL statement again.