The network agent also collects various TCP socket metrics at the host level for the local node on which it is installed. These metrics are useful for monitoring the TCP stack health, collisions, and errors on all the physical interfaces, and CPU consumption of the network agent.
TCP Socket Metrics in the Network Dashboard
These metrics measure the overall TCP socket health of an individual node. To view these metrics, select a node: go to the Network Dashboard, click on a tier, go to the Nodes tab, and select the node of interest. The TCP Socket metrics for the selected node appear on the right side of the dashboard.
Metric
Description
Default Monitoring Mode
Established
The number of sockets that are in the established state.
KPI
Embryonic
The number of sockets that are in the process of being set up.
KPI
Wait
The number of sockets that are unavailable for new connections because they are in the process of being closed. When a socket sends a Connection Close (FIN) message, the socket goes through three states:
FIN_WAIT1 – Socket has sent the FIN and is waiting for an ACK from the remote socket.
FIN_WAIT2 – Socket receives ACK, waits for FIN (final connection close) from the remote socket.
TIME_WAIT – The socket stays open to handle any packets for the TCP connection that might still be in the network. The TIME_WAIT state can last from 1 to 4 minutes, depending on the implementation.
A WAIT socket can result in significant delays and/or errors for the application or service that relies on that socket. A lot of simultaneous WAIT sockets can prevent applications and services from creating new connections.
KPI
Time Wait
In most cases, the FIN_WAIT states are relatively short (in milliseconds) while the TIME_WAIT state is relatively long (1 to 4 minutes). This means that the Wait and Time Wait metrics should be fairly similar. Significant differences indicate a significant network or other delays between the selected node and another host.
KPI
Interface Metrics in the Metric Browser
These metrics measure the number of collisions and errors on the physical interfaces of the node. To view these metrics, drill down in the Metric Browser as follows:
Application Infrastructure Performance >
<tier-name> >
Individual Nodes >
<node-name> >
Host >
Interface
Metric Name
Description / Notes
Monitoring Mode
# Collisions
Number of collision errors seen on all the interfaces, when an interface was blocked from sending a frame due to ethernet collision detection.
KPI
# Receive Drops
Number of packets dropped while receiving by all interfaces on the host
KPI
# Receive Errors
Number of collision errors seen on all the interfaces, when an interface could not receive a frame due to a physical problem on the interface or the connected cable.
KPI
# Transmit Drops
Number of packets dropped while transmitting by all interfaces on the host
KPI
# Transmit Errors
Number of collision errors seen on all the interfaces, when an interface could not send a frame due to a physical problem on the interface or the connected cable.
KPI
# Total Errors
The total number of ethernet and/or physical errors detected on the interface.