Specifications

Chapter 5: Configuring Equalizer Operation
44 Equalizer Installation and Administration Guide
• send buffer applies to L7 clusters and is the amount of memory in kilobytes reserved by each
L7 proxy process to store outgoing data before it is placed on the network interface.
• receive buffer applies to L7 clusters and is the amount of memory in kilobytes reserved by
each L7 proxy process to store data that has been received on an interface before it is
processed by an L7 proxy process.
• connect timeout applies to L7 clusters and is the time in seconds that Equalizer waits for a
server to respond to a connection request.
• client timeout applies to L7 clusters and is the time in seconds that Equalizer waits before
closing an idle client connection.
• server timeout applies to L7 clusters and is the time in seconds that Equalizer waits before
closing an idle server connection.
• probe interval is the target time in seconds at which the internal load balanci
ng
daemon checks the internal server status list that shows the resul
t of server probes (performed
by the probe daemon). If the list shows the server as down after strikeout threshhold
attempts to read the table, then the server is marked down. This value is solely a target; the
monitoring process adjusts itself based on load. The default value is 20
seconds.
Also see probe delay, below
.
• probe timeout is
the time in seconds that the probe daemon waits for a response once a TCP
or ACV probe has been issued.
• strikeout threshold is the number of failures (strikes) to respond to a TCP or ACV probe
before a server is declared down.
• log hours is the target number of hours of plot log data to retain. A zero in this field allots the
number of hours based on available memory. Note that the number of hours of log data
retained is limited by the amount of memory and disk space. If you define a large number of
clusters and servers, this will limit the amount of time over which log data can be retained on
Equalizer. For example, a system with 10 clusters each with 10 servers might only be able to
retain about 4 hours of log data.
• cycle time is time in seconds for the master daemon to make one pass through all of the
clusters. This value should not be modified unless recommended by Coyote Point Support.
• probe delay is the time in seconds (default is 10) between successive probes of servers by the
probe daemon. If a server fails to respond to a probe, the probe daemon marks it as down in
its internal server status table. This applies to both TCP probes (always performed by
Equalizer) and ACV probes (if enabled). You can override this value for each cluster.
• idle timeout applies to L4 clusters and is the time in seconds before reclaiming idle Layer 4
connection records.
• stale timeout the length of time that a partially open or closed connection is maintained; see
“Managing Stale Connections” on page 53.
• sticky netmask enables sticky network aggregation for a subnet (all the connections coming
from a particular subnet are directed to the same server in the cluster). See “Enabling Sticky
Network Aggregation” on page 54.
• command, from, to, and
subject enab
le
ev
ent hand
lin
g on the Equalizer;
see “Configuring
Cu
stom Event Handling” on page 114, in Chapter 7, for more information about these
parameters.