When configuring F5 Load Balancer, one crucial setting that impacts how the system manages traffic when a backend service becomes unavailable is Action on Service Down. This setting determines how the load balancer should respond when the target pool member is marked as down. In this blog post, we’ll dive into each of the options available and provide recommendations on when to use them.
Let’s take a closer look at each option for Action on Service Down and the scenarios in which they may be recommended.
Description: With this option, the system will not terminate existing connections to the pool member that has become unavailable. However, it won’t send any new traffic to that member.
Recommendation: This option can be used when it is essential to allow the existing connections to complete their work even if the pool member is no longer available for new connections. It is least disruptive to current traffic but doesn’t actively manage the unavailability situation.
Description: If no pool members are available, the system will reset and clear active connections, sending a TCP reset (RST) or Internet Control Message Protocol (ICMP) message. If there are pool members available, it will reset and clear active connections but route new connections to the available pool member without sending RST or ICMP messages.
Recommendation: Use this option when it’s important to quickly terminate connections to unavailable members and reroute new connections to available members. This can be particularly useful in cases where high availability and rapid failure recovery are critical.
Description: The system silently drops the connection without informing the client that the pool member is unavailable. This means no TCP RST or ICMP message is sent back to the client.
Recommendation: This option might be used in scenarios where you do not want to provide feedback to the client about the state of the connection. However, it can lead to timeouts and is generally not recommended for most use cases.
Description: If a pool member is marked down, the system will try to move existing connections to an alternative available pool member.
Recommendation: This option is useful in scenarios where it’s crucial to not drop any connections and ensure that the traffic is managed efficiently even if a pool member goes down. It is recommended for use cases where maintaining service continuity is critical.
In choosing the appropriate Action on Service Down, consider the nature of your application and the importance of availability, connection continuity, and how your application handles connection resets or failures. The Reject and Reselect options are generally more active in handling failures, while None and Drop are more passive.