What the node-to-external FPolicy server communication process is
To properly plan your FPolicy configuration, you should understand what the node-to-external FPolicy server communication process is.
Every node that participates on each storage virtual machine (SVM) initiates a connection to an external FPolicy server (FPolicy server) using TCP/IP. Connections to the FPolicy servers are set up using node data LIFs; therefore, a participating node can set up a connection only if the node has an operational data LIF for the SVM.
Each FPolicy process on participating nodes attempts to establish a connection with the FPolicy server when the policy is enabled. It uses the IP address and port of the FPolicy external engine specified in the policy configuration.
The connection establishes a control channel from each of the nodes participating on each SVM to the FPolicy server through the data LIF. In addition, if IPv4 and IPv6 data LIF addresses are present on the same participating node, FPolicy attempts to establish connections for both IPv4 and IPv6. Therefore, in a scenario where the SVM extends over multiple nodes or if both IPv4 and IPv6 addresses are present, the FPolicy server must be ready for multiple control channel setup requests from the cluster after the FPolicy policy is enabled on the SVM.
For example, if a cluster has three nodes—Node1, Node2, and Node3—and SVM data LIFs are spread across only Node2 and Node3, control channels are initiated only from Node2 and Node3, irrespective of the distribution of data volumes. Say that Node2 has two data LIFs—LIF1 and LIF2—that belong to the SVM and that the initial connection is from LIF1. If LIF1 fails, FPolicy attempts to establish a control channel from LIF2.
How FPolicy manages external communication during LIF migration or failover
Data LIFs can be migrated to data ports in the same node or to data ports on a remote node.
When a data LIF fails over or is migrated, a new control channel connection is made to the FPolicy server. FPolicy can then retry SMB and NFS client requests that timed out, with the result that new notifications are sent to the external FPolicy servers. The node rejects FPolicy server responses to original, timed-out SMB and NFS requests.
How FPolicy manages external communication during node failover
If the cluster node that hosts the data ports used for FPolicy communication fails, ONTAP breaks the connection between the FPolicy server and the node.
The impact of cluster failover to the FPolicy server can be mitigated by configuring the LIF manager to migrate the data port used in FPolicy communication to another active node. After the migration is complete, a new connection is established using the new data port.
If the LIF manager is not configured to migrate the data port, the FPolicy server must wait for the failed node to come up. After the node is up, a new connection is initiated from that node with a new Session ID.
The FPolicy server detects broken connections with the keep-alive protocol message. The timeout for purging the session ID is determined when configuring FPolicy. The default keep-alive timeout is two minutes.