NSK Agent Disconnects with U02000076 and U02000010 Errors Due to Network Idle Timeouts

Products

• CA Automic Workload Automation - Automation Engine

Issue/Introduction

NSK agents intermittently disconnect from the Communication Process (CP) after periods of inactivity, eventually failing to reconnect and entering an "ended_lost" status.․․․․

ERROR MESSAGE:

"U02000076 Connection to CP Server '[system-id]' closed." "U02000010 Connection to Server '[hostname]:<CP_Port#>' terminated." "AWAITIO-Error: 4127 on [IP-address]:<CP_Port#>"

 

SYMPTOMS:

• Agent logs show frequent connection terminations exactly every 11 minutes (660 seconds).

• The agent closes the connection with "Server does not respond anymore, will be closed."

• Network packet captures (PCAPs) show silence for approximately 5-6 minutes followed by a new SYN from the agent.

• No TCP RST or FIN packets are observed during the silent drop.

• The status changes to "ended_lost" after several days of failed reconnection attempts.

 

CONTEXT: Occurs specifically when no jobs are running for a prolonged period. Disconnects do not occur if a job runs at least every 3 minutes.

IMPACT: Automated processing on the HP NonStop platform is interrupted, requiring manual intervention to restore the agent connection.

 

Environment

OS: HP NonStop NSK L24.08.1

Agent: NSK Agent 24.4.0+build.1

AE Server Version: 24.4.1+build.1750690770737

AE Server OS: Windows Server 2022 

Network: Stateful routing boundary between Agent and AE

 

Cause

Intermediary network devices (firewalls or L3 switches) silently drop idle TCP sessions from their state tables after a common 5-minute timeout threshold. Because the agent's default internal timeout/check interval is set higher (660 seconds), it remains unaware that the connection was severed at the 5-minute mark until its own 11-minute timer expires, at which point it logs the U02000076 and U02000010 errors.

RELATED RESOURCES:

• KB-391956: NSK agent ignores CP_LIST entries

• KB-281079: NSK AWAITIO-Error

• KB-88825: What is Keep Alive and how does it work?

 

Resolution

PREREQUISITES:

• Access to the NSK Agent INI file.

• Authority to modify TCP/IP stack parameters on the HP NonStop OS.

 

STEPS:

1. UPDATE AGENT CONFIGURATION: Ensure the agent is permitted to use keep-alives.

   Action: Add or modify the following parameter in the [TCP/IP] section of the agent INI file: tcp_KeepAlive=Y

2. TUNE OS TCP KEEPALIVE PARAMETERS: Unlike Linux/UNIX agents, the NSK agent relies on the host OS's global TCP/IP stack to determine how often keep-alive packets are sent. Ensure the host OS stack timer values are shorter than the network's idle timeout (typically 5 minutes / 300 seconds).

   Action: Ask the NonStop team to review and tune the OS parameters. For example:

   • Idle time before first keepalive probe (e.g., 120 seconds).

   • Interval between keepalive probes (e.g., 30 seconds).

   • Number of retry probes before session closure (e.g., 3 to 5).

   EXPECTED: OS-level keep-alives will prevent silent session drops by intermediary stateful devices.

3. SUBMIT ENHANCEMENT REQUEST (RFE): Certification for NSK L24.x.

   Action: Submit an RFE via the Broadcom Customer Portal to get the specific NSK OS version officially certified.

   NOTE: NSK L24.x is currently outside the supported certification matrix for AE 24.4.1.