snmpwalk may fail with error Timeout: No Response from "hostname" when there are more than 1000 TCP or UDP connections (Doc ID 1357800.1)

Last updated on DECEMBER 08, 2022

Applies to:

Symptoms

The Solaris 10 or Solaris 11 system would be running the Solaris SNMP Agent, SMA, or a public domain version of Net-SNMP when this issue would occur.
When issuing a snmpwalk command to query the the system the snmpwalk may fail with the following error:
Timeout: No Response from "hostname"

This can occur when there are more than 1000 TCP connections, in any number of states, present in the system's TCP connection table when the command is executed. In most cases of this error it will take 8000 to 15000 connections to trigger the failure.
This mib cache size limitation can also occur with UDP connections.

In other circumstances it be may observed that Solaris 10's SMA agent process, snmpd, consumes a large amount of CPU when there are 10000 to 15000 connections for a sustained period of time or may core dump with a SIGSEGV.

One method of determining the number of TCP connections currently open on the Solaris 10 system would be to run a series of netstat -an commands, note the number of connections in each state, and then totaling the values for a overall number.

netstat-an.out | grep LISTEN | wc -l
netstat-an.out | grep BOUND | wc -l
netstat-an.out | grep ESTABLISHED | wc -l
netstat-an.out | grep CLOSE_WAIT | wc -l
netstat-an.out | grep TIME_WAIT | wc -l

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.