The Failure Detection Protocol monitors the core group network
connections that the Discovery Protocol establishes, and notifies the Discovery
Protocol if a connection failure occurs.
Before you begin
- Understand the concepts that are described in Core group Failure Detection Protocol
.
- Check your operating system settings that are relevant to TCP/IP socket
closing events.
- Determine your failure detection goals and which settings must change
to accomplish these goals.
About this task
You might want to perform this task if:
- You want to change the failover characteristics of your system.
- Your core groups are large and analysis indicates excessive CPU usage
is spent monitoring heartbeats.
To change the settings for the Failure Detection Protocol:
Procedure
- In the administrative console, click Servers > Core groups >
Core group settings > core_group_name.
- Under Additional Properties, click Custom Properties.
- Change the values specified for the IBM_CS_FD_PERIOD_SECS custom
property. This property specifies the time interval, in seconds,
between consecutive heartbeats. The default value for this property is 30
seconds.
If the IBM_CS_FD_PERIOD_SECS property already exists, click on
the property name, and in the Value field, specify the length of time, in
seconds, that you want the Failure Detection Protocol to wait between consecutive
heartbeats.
If this property does not already exist, click New and
create it:
- In the Name field, specify IBM_CS_FD_PERIOD_SECS.
- In the Value field specify the length of time, in seconds, that
you want the Failure Detection Protocol to wait between consecutive heartbeats.
- Change the values that are specified for the IBM_CS_FD_CONSECUTIVE_MISSED
custom property. This property specifies the consecutive number
of heartbeats that must be missed before the protocol assumes that the core
group member has failed. The default value for this property is 6.
If
the IBM_CS_FD_CONSECUTIVE_MISSED properties already exists, click on the property
name, and in the Value field, specify the number of heartbeats that must be
missed before the Failure Detection Protocol assumes the core group member
failed.
If this property does not already exist, click New and
create it:
- In the Name field, specify IBM_CS_FD_CONSECUTIVE_MISSED.
- In the Value field specify the number of heartbeats that must
be missed before the Failure Detection Protocol assumes that the core group
member failed.
- Click OK, and then click Save to
save your changes.
- Select Synchronize changes with nodes,
and then click Save again.
- Restart all of the members of the core group.
Results
After the servers restart, the core group members all run with the
new Failure Detection Protocol settings.