F5 Updated F5CAB3 Exam Questions and Answers by fynn

This scenario describes a classic case of asymmetric routing in a "one-arm" or non-gateway deployment. When a BIG-IP system is configured for SSL offloading, the following traffic flow occurs:

Client-Side: The client establishes a successful SSL/TLS handshake with the Virtual Server. This explains why the user can "successfully establish an SSL connection."

Server-Side: The BIG-IP decrypts the traffic and forwards it as plain HTTP to the backend server. The packet trace confirms the server receives the HTTP GET request and responds with the content.

The Routing Failure: By default, the BIG-IP system preserves the client's original source IP address. If the backend server's default gateway is not the BIG-IP system (or if the server is on the same subnet as the client), the server will attempt to send the response directly back to the client's IP address, bypassing the BIG-IP.

Stateful Drop: Because the BIG-IP is a Full Proxy, it expects the response to return through its own internal state table to be encrypted and sent back to the client. Since the response bypasses the BIG-IP, the BIG-IP connection eventually times out, and the client receives no data despite the server having sent it.

Solution (SNAT): Enabling Secure Network Address Translation (SNAT), specifically SNAT Auto Map, ensures that the BIG-IP replaces the client's source IP with its own internal self-IP before sending the request to the server. This forces the server to send the response back to the BIG-IP, allowing the BIG-IP to complete the transaction and deliver the content to the user.

When a pool contains servers with heterogeneous hardware capabilities (differing CPU, RAM, or connection limits), a static load balancing method like Round Robin is ineffective because it distributes requests equally, regardless of the server's capacity. To optimize traffic distribution for newer, more powerful servers, a dynamic or weighted method is required.

Weighted Least Connections (member): This is the ideal method for this scenario. It combines two factors:

Least Connections: It first checks the current active connection count to ensure traffic goes to the least busy server.

Weight (Ratio): It allows the administrator to assign a "Ratio" value to each pool member. Newer servers can be assigned a higher ratio (e.g., 3) while older servers are assigned a lower ratio (e.g., 1). The BIG-IP system uses these weights to disproportionately favor the newer servers even when connection counts are similar.

Why other options are incorrect:

Global Availability: This is primarily a GSLB (Global Server Load Balancing) or specific LTM priority group concept where traffic is sent to the first available member in a list until it fails, then moves to the next. It does not load balance based on capacity.

Round Robin: This passes each new connection request to the next server in line, treating the old and new servers exactly the same.

Least Connections (member): While this sends traffic to the server with the fewest active connections, it assumes all servers are equal. If an old server and a new server both have 10 connections, they are treated as equally capable of taking the 11th, which is not true in this scenario.

From the exhibit, the pool contains different applications on different service ports (for example, HTTP/80, FTP/21, HTTPS/443, SSH/22). To mark pool members correctly, BIG-IP must be able to verify the actual service running on each member’s port.

In BIG-IP Administration: Data Plane Configuration, monitor behavior is described as follows:

When multiple monitors are assigned to a pool, the Availability Requirement controls how monitor results are evaluated:

At least one = the pool member is marked up if any one of the assigned monitors succeeds.

All = the pool member is marked up only if every assigned monitor succeeds.

For pools containing members with different services/ports, using All can incorrectly mark members down because monitors intended for other services will fail on the wrong port.

Why C is correct:

Assigning HTTPS, HTTP, FTP, and SSH covers the actual services shown in the pool.

Setting the Availability Requirement to at least one ensures that each pool member is considered available when its appropriate service monitor succeeds, without being forced to pass unrelated service monitors.

Why the other options are incorrect:

A / D (Availability Requirement = all): would cause members to be marked down when unrelated monitors fail (e.g., SSH monitor against an HTTP member).

B (includes ICMP): ICMP can indicate the host is reachable even if the application service is down, which does not “accurately” reflect service availability.

Therefore, the best choice is HTTPS, HTTP, FTP, and SSH with Availability Requirement of at least one health monitor.

Similar to the logic required for managing multi-service backend environments, the issue described—where servers hosting multiple protocols like FTP and HTTP are experiencing uneven distribution—stems from the BIG-IP's default behavior of treating each pool independently. If the administrator uses a member-based load balancing method, the BIG-IP distributes HTTP traffic regardless of how much FTP traffic that same physical server is currently processing.

To resolve this, the administrator must utilize the Least Connections (Node) method. By switching both the HTTP and FTP pools to this algorithm, the BIG-IP begins to make load balancing decisions based on the total combined connection count for the IP address of each server. When a new HTTP request arrives, the BIG-IP checks which server has the fewest total connections (including existing FTP sessions). This prevents a server that is already busy with long-lived FTP transfers from being overwhelmed by a sudden burst of HTTP requests.

Ratio methods (Options A and C) are static and rely on the administrator manually assigning weights to servers based on their perceived capacity; they do not adapt to real-time fluctuations in traffic volume across different pools. Least Connections (Member) (Option B) remains blind to the "cross-pool" traffic on the same hardware. Only the Node-based Least Connections approach provides the global visibility necessary to "even out" the total resource utilization across servers supporting multiple distinct applications.