Signiant Support

Outbound traffic on clustered agent or HA manager is sending from active node's interface Print


Problem: 


In a clustered environment, either a clustered agent or a manager using the High Availability option, outbound traffic from the cluster is originating from the active node's interface instead of the cluster's virtual interface.

Resolution:


The agents needs to be configured specifically to use the cluster's virtual IP address.

For HA managers, or for clustered agents that share a common configuration, follow these steps:
Log onto the manager GUI as an admin level account.
From Administration / Agents / Configure, find the agent you want.
Click on the Configure link in the top menu.
Change the value of the IP Interface field (about 3/4 down the list of items) to the virtual IP (VIP) address for your cluster.
Log onto the agent as a root or administrator account Restart the Signiant UDP Relay service:

Windows:  use the control panel or type "net stop ddsursrv && net start ddsursrv" on the command line.
Linux / Mac / Unix:  run /usr/signiant/dds/init/siginit restart sigur (assuming a default install location).

This is the only service that needs to be restarted -- all others take the change immediately.
For clustered agents that do not share a common configuration file, follow the above steps for the active node.  When this is complete, follow these steps:
Log onto the passive agent, via the console, remote desktop or through SSH.
Locate the agent's configuration file.

Windows:  C:\Program Files\Signiant\Mobilize\bin\dds.cfg in a default installation
Linux  / Mac / Unix:  /etc/dds.conf in a default install

Edit the file in a text editor.
Add the following line:

IP interface is xxx.xxx.xxx.xxx

Where xxx.xxx.xxx.xxx is the VIP address for your cluster.  If this line already exists, edit the IP address specified on this line to the cluster's VIP address.
Note:  This configuration only works in an active-passive cluster configuration where the agent is running on one node at a time.  If your cluster is an active-active cluster, both nodes will be competing to bind to the VIP address, resulting in unpredictable agent behaviour.

Additional Information
See the KB article titled "Configure agents on RedHat clusters with shared storage" for an example of how to configure agents on Red Hat Clusters in an active-passive mode using a shared configuration.