Tcp Probe Aggregation - Coyote Point Systems E350GX Installation And Administration Manual

Appendix B: Timeout Configuration
The figure below shows the relationship between the
does not respond to a High Level Probe.
Figure 63 Unsuccessful probe timeout timeline
In the figure above, a High Level Probe (HLP) is sent to a server, which does not respond before the
elapses. Equalizer marks the server as failing and sends two additional HLP probes (the default value of
is 3).
threshold
The specifies the time between these additional probes. If the server does not respond to either of
probe interval
these probes, it is marked down.
Note that the time periods between probes specified by the
times. 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 addition, settings below 5 have the same effect as a setting of 5, since the Equalizer probe
daemon cycles through the server probes every 5 seconds.
The range of values for each HLP parameter is shown in the table below. These apply to TCP and ACV probes only:
Parameter
probe timeout
probe delay
probe interval
strikeout threshold

TCP Probe Aggregation

If a server is defined in more than one cluster and ACV probing is not enabled on any of the clusters to which the
server belongs, then probes for that server are aggregated -- meaning, Equalizer only sends the server one TCP probe
during each probe cycle, instead of sending one probe for each cluster. This reduces redundant probing.
Once ACV probing is enabled in a cluster, however, probe aggregation is disabled for all the servers in that cluster,
and for all instances of these servers in other clusters.
For example, assume we have three clusters:
clusters. If ACV probing is not enabled in any of the three clusters, then
probe cycle, and the probe status will be reflected in all the clusters. Let's say we then enable ACV probing in
cluster . Thereafter,
A sv01
times, once for each cluster
It is possible, therefore, that defining a single server in many clusters and subsequently enabling ACV probes in any
one of these clusters may result in a noticeable increase in probe traffic on the network.
268
Minimum
1.0
0.0
0.5
, , and
A B
will be probed independently for each cluster -- that is, Equalizer will probe
, , and .
A B C
and
probe timeout probe interval
and
probe interval probe delay
Default Maximum
10.0
10.0
20.0
1 3
. Let's say the same server,
C
will be probed only once during each
sv01
Equalizer Installation and Administration Guide
parameters when a server
server timeout
strikeout
values are minimum
Units
60.0 seconds
60.0 seconds
25.0 seconds
6 integer
, is defined in all three
sv01
sv01
three