Coyote Point Systems E350GX, Installation And Administration Manual

Page: 271 / 344

Server Health Check Probes and Timeouts
Equalizer Installation and Administration Guide 267
examines the first 1024 characters for the ACV response string: Since this is done as part of the same
connection as the TCP probes, the same
probe timeout
period also applies to the ACV probe (i.e., the
probe timeout
timer is not reset):
• If the server response is received before the
probe timeout
expires and contains the ACV
response string, it is marked up, and Equalizer waits for the
probe delay
period before it starts
the HLP probing process again at
Step 1
.
• Otherwise, if the server does not respond to the ACV probe before the
probe timeout
expires, or
if the server responds but its response does not contain the ACV response string, then Equalizer
marks the server
failing and starts sending strikeout probes; go to
Step 3
.
3. This step is only performed when a server does not respond to a TCP or ACV probe within the
probe timeout
,
and is marked failing. Before marking a failing server as down, Equalizer sends additional probes until the
number of probes sent equals the value of the
strikeout threshold
parameter. The time between these additional
probes is specified by the
probe interval
parameter:
• If the failing server does not respond to any of the strikeout probes, it is marked down . Equalizer then
continues sending TCP probes to the server using
probe interval
as the minimum delay between
probes. If a response is ever received, Equalizer marks the server up and waits for the
probe delay
period before it starts the probing process again at
Step 1
.
• If a failing server responds to one of the strikeout probes, Equalizer marks the server up and waits for
the
probe delay
period before it starts the probing process again at
Step 1
.
The following figure shows the relationship between the
probe timeout
and
probe delay
parameters in a successful
probing sequence.
Figure 62 Successful probe sequence timeline
A server must respond to both TCP and ACV probes before the
probe timeout
elapses. Therefore, when ACV
probes are enabled and probe failure messages are showing up in the log, you may need to increase the
probe
timeout
so that the server has sufficient time to respond to both probes.
Assuming that a server responds successfully to TCP and ACV probes, Equalizer then waits for the
probe delay
time period before it sends the next TCP probe to the same server. (Note that the
probe delay
value is the minimum
time between successful probes; the observed time may be longer for large configurations with many servers, during
periods of high traffic, or due to Equalizer adjusting the delay internally to prevent server probes from consuming
too much bandwidth on the network interface.)
In a network configuration where there is high latency between server probes and responses, the probe mechanism
may falsely report that a server is down; this is indicated by messages in the event log indicating that a server is
down and then comes back up again after a short period of time. In such cases you may need to increase the
probe
timeout
or the
probe delay
parameters (or both) to reduce the number of false server down conditions reported by
the probing mechanism.