Smarts IP: AgentNotResponding event is not generated immediately when a device's SNMP Agent stops responding; How long should it take?

Products

• VMware Smart Assurance

Issue/Introduction

Symptoms:


How long should it take the Smarts IP AgentNotResponding event to be generated after a device's SNMP Agent stops responding?

Smarts IP AgentNotResponding event is not generated immediately when a device's SNMP Agent stops responding

Environment

VMware Smart Assurance - SMARTS

Cause

The AgentNotResponding event is not generated immediately, as Smarts IP needs to ensure that the SNMP agent has really stopped and it is not just a case of dropped SNMP packets due to temporary network congestion issues and so on.

Resolution

There is not an exact timeline on when the AgentNotResponding event will be generated after the SNMPAgent actually stops responding as it depends on the current state of the polling cycle. This functions as follows:

1. When the SNMP agent on the device stops responding, the SNMPStatus attribute goes to TIMEDOUT to indicate that the agent is not responding to SNMP requests but the host system is still responding to ICMP pings.
2. The Status attribute, which shows the current status of the SNMP agent, will go to UNRESPONSIVE and the TimeOutTicker attribute will start counting up (after roughly one polling cycle of failed SNMP requests).
3. The AgentNotResponding event will then be generated based on the following logic:
   
   AgentNotResponding
   = (SNMPStatus != SNMPAgent::OK &&
   SNMPStatus != SNMPAgent::UNKNOWN &&
   TimeOutTicker >
   polling_frequency(sysUpTime) * AgentNotRespondingTimeOut)
   else FALSE;

There are three variables associated in the above logic:

• SNMPStatus: Polled using the sysUpTime OID
• TimeOutTicker: Starts counting up after at least one polling cycle of failed SNMP requests and it displays the time since IP detected that the SNMPAgent stopped responding.
• AgentNotRespondingTimeOut: This is a multiplier for the PollingInterval (polling_frequency returns the time between successive polls), by default AgentNotRespondingTimeout has a value of 3.

Essentially, the above logic indicates the AgentNotResponding event will be generated when the TimeOutTicker is greater than the polling frequency * AgentNotRespondingTimeOut, with default settings of 240 * 3.  So in a default configuration, Smarts IP needs at least one polling cycle where there was no response from SNMP requests for the TimeOutTicker to start counting, and then 3 more polling cycles for the multiplication factor to be exceeded by the TimeOutTicker. In total this will be roughly 4 polling cycles. This is when the AgentNotResponding event will be generated by Smarts IP and displayed in SAM.