HP e-Commerce / XML director server appliance sa8250 - Users Guide

C H A P T E R 7 Using SNMP
265
Server Availability (Ping)
NOTE: The Intelligent
Resource Verification
(IRV) CLI command is
config irv <ping-
interval>
and the
default ping-interval is
zero. To make the
serverPingTable active,
ensure that the ping-
interval is NOT set to
zero.
The
serverPingTable can be used to monitor server availability.
If a server is responding to periodic ping requests from the SA8250,
then its state is marked as
responding. Otherwise, the server is
marked as
notResponding. Whenever a server’s Ping state toggles,
the trap
serverPingNormal or serverPingAlert is sent for
that server.
Server TCP Connection
The performance of TCP connections between the SA8250 and
servers are monitored. This performance data is stored in the
serverTcpTable. For each configured server mapping used for
load balancing a service (VIP, PORT), the following data is
maintained in the
serverTcpTable:
• state, up or down (
serverState)
• how long this server has been up (
serverUpTime)
• response time (
serverRspTm)
• number of established TCP connection instances
(
serverConnCnt)
• TCP connections established per second (
serverCps)
Trap thresholds are available in the MIB for
serverRspTm,
serverConnCnt and serverCps. Different threshold settings can
be configured for each server TCP connection in
serverTcpTable.
Trap thresholds for server response time can be configured so that a
trap is sent if the response time reaches a specified value. When
serverRspTm reaches serverRspTmHiWater, a
serverRspTmAlert trap is sent. While in alert, if serverRspTm
dips to
serverRspTmLoWater, a serverRspTmNormal trap is
sent. High- and low- water thresholds provide hysteresis and prevent
the spurious generation of traps. If the high-water threshold is set to
0, no traps are sent.
Trap thresholds for server connection count can be configured so that
a trap is sent if the connection count reaches a specified value. The
serverConnCntAlert and serverConnCntNormal traps and
applicable thresholds work similarly for server response time.