Case Studies

This section explores the failover features available on PIX Firewall in greater detail (the discussion is based on information from the New Product Introduction training by PIX marketing team with modification to be able to easily comprehend).

The resiliency of the connections through the PIX firewall can be achieved with failover, which means that if one PIX failed, the other PIX still would be available to process the packets. Starting from PIX Version 7.0, PIX can be configured in one of the following two failover modes:

• Active/standby model

• Active/active model

Active/Standby Model

In this model, one PIX acts as the active PIX that processes the traffic at the time, and the other unit acts as standby. In the event of failure of the active unit, the standby unit becomes active and starts processing packets. If stateful failover is configured, the transition of connections from active to standby is very smooth. The new features added to Version 7.0 are as follows:

• Stateful failover for VPN traffic This release introduces stateful failover for VPN connections. All security association (SA) state information and key material is synchronized automatically between the failover pair members, and this provides a highly resilient VPN solution.

• Non-Stop Online Software Upgrades This version allows you to perform software upgrades of failover pairs without affecting network uptime or connections flowing through the units. This is because of the ability being introduced in this version to perform inter-version state sharing between PIX failover pairs. This allows you to perform software upgrades to maintenance releases (for example, 7.0(1) upgrading to 7.0(2)) without affecting traffic flowing through the pair. There is no impact in both active/standby failover environments and active/active environments.

Active/Active Model

The active/standby model is not the best way to use the resources. Ideally, both of the units in the failover should be used, and act as backups for each other in the event of failure. This is exactly what active/active failover can provide, as introduced in Version 7.0. In this model, both units can actively process firewall traffic while at the same time serving as backups for each other. In PIX Version 7.0, only a two-node setup is possible, in which each device handles 50 percent of the traffic. When one unit fails, the remaining unit takes over and processes 100 percent of the traffic.

Remember that in this model, PIX itself does not perform the load balancing of the traffic. So, it is your responsibility to engineer the network to route 50 percent of the traffic to each unit. This can be accomplished either by static routes or dynamic routing protocols on both upstream and downstream routers.

A new command, failover groups, is being introduced to support active/active failover. Conceptually it is very similar to the Cisco's Multi-Group Hot Standby Router Protocol (HSRP) or Virtual Router Redundancy Protocol (VRRP). As mentioned before, PIX 7.0 supports a two-node active/active failover configuration with two failover groups.

The active/active failover feature leverages the Security Context feature. Figure 3-10 shows a typical two-node cluster. On unit A, ctx1 is active and ctx2 is standby. Unit B has ctx2 as active and ctx1 is standby. If unit A fails, unit B will have both contexts active and will process 100 percent of the firewall traffic.

Figure 3-10. Active/Active FO setup

Each failover group contains separate state machines to keep track of the failover state. The command failover active group <group#> can be used to change the active state of a failover group.

Hardware and License Requirements

You must have the following minimum requirement for hardware and licensing to configure active/active failover:

• Both units need to have the same hardware configuration.

• Both units must have an unrestricted (UR) License, or the primary unit must have a UR license and the secondary unit must have an active/active failover-only license.

System and User Failover Group

To support active/active failover, PIX 7.0 failover added support for failover groups. Each failover group has its own state machine and can switch over independently. There are two types of failover group: system failover group and user failover group. The concept is similar to the system and user context.

The system group is used internally by the failover process. Its main purpose is to allow the failover process to manage the unit-wide activities under the failover group scheme. These activities include: unit health monitoring, failover command interface health monitoring, running config synchronizing, and so on. There is only one system failover group per unit, and it is created automatically when the user enables the failover. The system context is bound to the system failover group.

User failover groups are used to manage the user contexts under the active/active failover scheme. Figure 3-10 shows a simple example of two user contexts that are bound to two user failover groups.

Note

PIX Firewall Version 7.0 supports a maximum of two (user) failover groups. However, there can be more than two user contexts configured in the system.

The failover-group subcommand under the context command is used to bind a user context to a failover group. The failover group 1 is the default failover group. If a user context is not bound to a failover group through the command interface, it is bound to failover group 1 by default. Failover group 1 must be created first and must be the last group to be removed. If a context is bound to a failover group, the failover group cannot be deleted unless the binding is removed.

Initialization, Configuration Synchronization/Command Replication

When a unit boots up, the system failover group will contact the active peer to obtain the running configuration. If both units boot up at the same time, the System failover group of the Primary unit will become active and synchronize its configuration to the Secondary unit. After configuration syncing has finished, the state machine of the user failover group will start running to elect the active unit for each group and start the Active/Active failover.

Even though both units could be actively processing user traffic, the command replication is uni-directional. The command will be replicated to the peer only if failover group 1 is in active state. Users will be warned if they try to enter a configuration command from the standby unit. This means that a user context can be in the active state, but the user will need to enter commands from the user context of the standby unit if it is bound to failover group 2.

Configuration Examples

Work through the steps that follow to configure Active/Active Failover:

Step 1.	Verify that both units have exactly the same hardware configuration and proper license.
Step 2.	If a unit is in single context mode, use the command: mode multiple to bring it to multiple security contexts mode.
Step 3.	Configure the basic failover parameters in the system context of the primary unit. Example 3-26 shows the configuration needed for active/active failover. Example 3-26. Active/Active Failover Setup PIX(config)# failover lan unit primary PIX(config)# failover lan interface folink ethernet2 PIX(config)# failover link stfo Ethernet3 PIX(config)# failover interface ip folink 1.1.1.1 255.255.255.0 standby 1.1.1.2 PIX(config)# failover interface ip stfo 2.2.2.1 255.255.255.0 standby 2.2.2.2 ! This command is optional but recommended PIX(config)# failover key cisco123 PIX(config)# failover lan enable PIX(config)#
Step 4.	Use the failover group command to configure a failover group. Example 3-27 shows how to configure two failover groups with the preemption option. Example 3-27. Failover Group Configuration ! Execute failover group command that will take you to the subcommand mode PIX(config)# failover group 1 ! Primary unit has higher priority PIX(config-failover)# primary ! Preempt peer if bootup as Standby PIX(config-failover)# preempt PIX(config-failover)# exit PIX(config)# failover group 2 ! Secondary unit has higher priority PIX(config-failover)# secondary PIX(config-failover)# preempt PIX(config-failover)# exit PIX(config)#
Step 5.	Bind the user contexts to the failover group. Assume there are two contexts, ctx1 and ctx2, in addition to admin. PIX(config)# context ctx1 PIX(config-context)# join-failover-group 1 PIX(config-context)# exit PIX(config)# context ctx2 PIX(config-context)# join-failover-group 2 PIX(config-context)# exit PIX(config)#
Step 6.	Type the failover command to enable active/active failover.
Step 7.	Configure the bootstrap failover configuration on the secondary unit as shown in Example 3-28. Example 3-28. The Initial Configuration on the Standby PIX PIX(config)# failover lan unit secondary PIX(config)# failover lan interface folink ethernet2 PIX(config)# failover link stfo Ethernet3 PIX(config)# failover interface ip folink 1.1.1.1 255.255.255.0 standby 1.1.1.2 PIX(config)# failover interface ip stfo 2.2.2.1 255.255.255.0 standby 2.2.2.2 ! This command is optional but recommended PIX(config)# failover key cisco123 PIX(config)# failover lan enable PIX(config)# failover
Step 8.	After failover is up, on the Primary unit (i.e., Active for failover group 1), issue the write memory command to save the configuration to flash.

Asymmetrical Routing Support

For better performance and reliability, you may have the network set up with redundant connections to the same Internet Service Provider (ISP) or two different ISPs as shown in Figure 3-11. This poses a problem for PIX, as the ISP may receive traffic from one PIX but might return traffic back to the other PIX. This is because these two PIX firewall units (or pairs) do not share session information. Therefore, while the return packet is allowed by the PIX that allowed the initial connection, it is denied by the other PIX. This is also how it works in Failover mode when it is deployed with Active/Standby Mode.

Figure 3-11. Asymmetric Routing Support with Active/Active FO Setup

The same problem can occur when running active/active failover. A unit may receive a packet that belongs to its peer. If this happens, the receiving unit will forward the packet back to its peer for processing. Stateful failover must be enabled to support asymmetric routing.

Figure 3-11 shows two units running Active/Active failover, where Unit 1 has context ctx1 active and Unit 2 has context ctx2 active. An inside host initiates a connection through context ctx1 of Unit 1 (solid line). Context ctx1 creates the connection, replicates the connection to Unit 2, and then forwards the packet. If the return packet is routed through context ctx2 of Unit 2, connection information already exists in Unit 2's connection table. But, because context ctx1 is not active in Unit 2, it forwards the packet back to Unit 1 (arrows).

Following are some of the restrictions and cautions you should remember when configuring asymmetric routing in Active/Active Failover mode.

• Remember that multiple context PIX does not support VPN, IPS, and other features.

• Shared interface setup requires NAT.

• Under the asymmetric routing A/A FO, packets are forwarded by Layer 2.

• The supported scenario is for outbound traffic being routed through one unit and inbound traffic being routed to a different unit for a given connection. The scenario where traffic traveling in the same direction for a connection gets routed to different units should not happen if routing is configured properly. The upstream or downstream router should set up the load-balancing policy such that routers are not performing per packet load balancing to the High Availability (HA) cluster.

To configure asymmetric routing support, you need to use asr-group command under the interfaces of contexts. Asymmetric routing support is needed for the outside interface only; you need to have the following commands configured on both ctx1 and ctx2:

PIX/ctx1(config)# interface Ethernet 0.30 PIX/ctx1(config-if)# asr-group 1 PIX/ctx2(config)# interface Ethernet 0.40 PIX/ctx2(config-if)# asr-group 1 

Troubleshooting Steps

Before working through some of the common scenarios, it would be helpful for you to examine the syslog messages that are shown on the secondary PIX firewall after you turn on logging on as shown in Example 3-29.

Example 3-29. Syslog on the Secondary PIX

PIX(config)# logging on PIX(config)# logging monitor 7 PIX(config)# 111008: User 'enable_15' executed the 'logging con 7' command.         Detected an Active mate.  Switching to Standby         Switching to Standby. 

After the units are synchronized with each other, you can find out the status of a unit on both the primary and secondary with the show failover command. Example 3-30 shows the output of the show failover command on the primary unit.

Example 3-30. Monitoring Failover Status on Primary PIX Firewall

PIX(config)# show failover Failover On Cable status: Normal Reconnect timeout 0:00:00 Poll frequency 5 seconds Last Failover at: 14:52:29 EST Wed Feb 9 2005         This host: Primary - Active                 Active time: 14805 (sec)                 Interface outside (192.168.1.1): Normal                 Interface inside (10.1.1.1): Normal                 Interface stateful (1.1.1.1): Normal         Other host: Secondary - Standby                 Active time: 250 (sec)                 Interface outside (192.168.1.2): Normal                 Interface inside (10.1.1.2): Normal                 Interface stateful (1.1.1.2): Normal Stateful Failover Logical Update Statistics         Link : stateful         Stateful Obj    xmit       xerr       rcv       rerr         General         34036      0          1054      0 PIX# 

Work through the following steps to troubleshoot the failover problem:

Step 1.	Level 1 syslog will give the reasons for a failover. So always check the syslog to determine the root cause. For example, if the switch port failed on the inside interface of Active Firewall, you would see the following message on the Primary (Active) firewall. 411002: Line protocol on Interface inside, changed state to down 105007: (Primary) Link status 'Down' on interface 1 104002: (Primary) Switching to STNDBYinterface check, mate is healthier Syslog from Secondary (Standby) Firewall will report the following message: 104001: (Secondary) Switching to ACTIVEmate want me Active
Step 2.	Execute show interface on both PIX firewalls to make sure they are up.
Step 3.	Test the connectivity by pinging to the Failover interface IP. Be sure to allow the ICMP for the interfaces.
Step 4.	If the primary and secondary are connected in two different switches, be sure that all VLANS are trunked between the switches.
Step 5.	Be sure to turn on dot1Q across the board on the switch.