Like Server Visibility agents, network agents support Dynamic Monitoring Mode (DMM). Rather than have all network agents report all metrics all the time, you can run each agent in one of three modes:
- KPI Mode – Network KPI metrics for all monitored objects (application flows, tiers, nodes, and network links) except individual connections
- Diagnostic Mode – Network KPI metrics for all monitored objects, including connections
- Advanced Diagnostic Mode – All network metrics for all monitored objects, including connections
Every Network Visibility metric has a default DMM class (KPI, Diagnostic, and Advanced Diagnostic). To see the DMM class for each metric, see Network Visibility Metrics.
All network agents run in KPI mode by default. The recommended workflow is to
- Run all network agents in KPI mode.
- When you see a performance issue on a specific node or network link, increase the metric level on the associated network agents to Diagnostic. Collect KPI metrics for the connections.
- Identify the connections with performance issues, based on the connection KPIs.
- To troubleshoot an individual connection, increase the metric level on the associated network agents to Advanced Diagnostic. Collect advanced metrics for the connection.
- When the issue is resolved, reset the agents back to KPI mode.
Changing the DMM on a Network Agent
- Click the gear icon ( ) in the top-right corner of the Controller page, choose AppDynamics Agents, and go to the Network Visibility Agents table.
- Select the agents of interest, right-click, and choose Change Dynamic Monitoring Mode.
The network agent retains its DMM setting even if the agent is stopped and restarted. See Managing Network Agents in the Controller.
Connection Diagnostics and Network Bottlenecks
Network agents can collect an extensive set of individual Connection metrics within a network link. Some TCP performance issues that Network Visibility can detect include:
- The data-receive window is too small (or zero) on the client or the server, which slows down the transfer of data.
- One or more servers are experiencing errors setting up or tearing down the connection for an individual TCP session.
- The client and server take a long time to set up an individual session.
- The application is using a lot of short-lived connections. TCP is most efficient when long, stable connections are used.
- Some TCP sessions have unusually high round-trip times (RTTs). When TCP is performing well, RTTs are stable and determined mostly by the network path between two nodes.
The network agent does not collect any Connection metrics in KPI mode (the default setting). You can change the Dynamic Monitoring Mode on a network agent when you need to diagnose a node or network path that is monitored by that agent.
After you complete the initial setup, you can set the Dynamic Monitoring Mode on individual network agents as needed. An example workflow might look like the following:
- The DevOps team for a large enterprise monitors its IT infrastructure using network agents on critical servers. All agents are initially set to KPI mode.
- The Network Dashboard shows a spike in latency on a network link between tier-A and tier-B.
- A DevOps team member does the following:
- Sets the network agent DMM on the tier-A and tier-B servers to Diagnostic.
- Collects KPI metrics for the Connections between the tier-A and tier-B nodes.
- Identifies one high-latency connection between nodes TA-N1 and TB-N3, based on the Connection KPIs. The KPIs for all other connections are within acceptable bounds.
- Sets the network agent DMM on TA-N1 and TB-N3 to Advanced Diagnostic and the DMM on all other agents to KPI.
- Notes a set of spikes in Nagle delays on this Connection. These spikes correspond to the latency spikes she noticed on the network link.
- Reconfigures TCP on the two nodes and monitors the Connection. The Nagle's and latency spikes no longer occur.
- Resets DMM on the TA-N1 and TB-N3 agents back to KPI.