HP Traffic Director sa7220 - Manual (Page 43)

C H A P T E R 2 Error Detection
31
Error
Detection The SA8220 is capable of recognizing and reacting to server error
conditions, detecting non-responsive (comatose) servers, and
directing traffic to alternate resources until the server is back in
operation. The SA8220 can also capture many HTTP errors before
they reach the client, and redirect the request to an alternate server.
Server Status Detection
The SA8220 uses multiple means to monitor the status of the
fulfillment servers. The “Intelligent Resource Verification” (IRV)
module periodically pings the servers to verify they are alive. The
SA8220 also monitors a “dup-syn” interval to calculate packet loss
rate.
Intelligent Resource Verification
When the IRV module pings a server and receives no response, it tries
to connect to each port on which the suspect server is configured to
listen. If the SA8220 itself does not receive a response from a given
port, then that server/port combination is declared dead. If the server
maintains network connectivity and responds positively to IRV
pings, but ports stop responding, then the dup-syn interval threshold
(described below) is used to decide if the server is declared dead.
Dup-syn Interval
The SA8220 dynamically calculates the threshold for the acceptable
number of dropped packets within a given interval. If at any time in
this interval the number of dropped packets exceeds this threshold,
the server is considered dead. After the specified time value has
expired the lost packet (or dup-syn) count is divided by two and the
time interval starts again. In this way, some history information is
kept between time intervals.
The dup-syn interval for this threshold is established with the dup-
syn CLI command, and ranges in value from 1000 to 2,147,483,647
microseconds. The default time interval value is 500,000
microseconds (one half second), which is appropriate for most
environments. By lowering or raising this value, you render the
SA8220 respectively less or more sensitive to dropped packets, and
less or more likely to declare a server dead. The volume of network
traffic must be taken into account when setting the dup-syn interval.
Higher volumes of traffic require a shorter dup-syn interval to avoid
mistakenly declaring a server dead due to network congestion.