gpssh commands timeout after ~30 seconds during long-running operations in Greenplum Database

Products

• VMware Tanzu Greenplum
• VMware Tanzu Greenplum / Gemfire
• VMware Tanzu Data Suite
• VMware Tanzu Data Suite

Issue/Introduction

You observe that gpssh sessions terminate after approximately 30 seconds when you execute long-running commands (for example, large pip install operations) across hosts in your Greenplum Database cluster. This behavior interrupts your administrative or deployment workflows.

Cause

This issue is not caused by gpssh itself. You are encountering this behavior due to underlying OpenSSH client and server configurations.

Since gpssh is a wrapper around SSH, it inherits SSH behavior, including connection timeout and keepalive handling. If you do not configure SSH keepalive parameters properly, idle or long-running sessions terminate prematurely.

Resolution

You cannot configure timeout settings directly in gpssh. To resolve this issue, you must tune SSH client and server keepalive settings to support long-running operations.

Update SSH configurations on all hosts in your cluster to prevent idle timeout during extended tasks.

Additional Recommendations

• You should review and tune the following parameters in sshd_config for high parallelism workloads:
  
  
  • MaxSessions
  • MaxStartups


• Ensure consistent SSH configuration across all nodes to avoid uneven behavior.

By aligning SSH keepalive and session settings with your workload requirements, you ensure that gpssh supports long-running operations without premature disconnections.