My previous lab session ended with a working IPsec site-to-site VPN between two standalone SRX firewalls. Then I decided to make it more complicated by adding chassis cluster on top. The idea is to have two physical SRX devices acting as one active/passive firewall. The IPsec configuration stays the same. If the primary node dies, the secondary picks up in under three seconds and nothing downstream notices.
How It Works
Two SRX nodes share one configuration file and one control plane. The active node (primary) handles all traffic while the standby mirrors its session state in real time: routing table, active TCP sessions, IPsec security associations. When the primary fails, the secondary already has everything it needs.
The IP addresses do not move the way you might expect. They live on reth interfaces, which are logical interfaces that span both nodes and follow whichever one is currently primary. On failover, the new primary sends a gratuitous ARP and traffic shifts.
Two physical cables make this work. The control link handles heartbeats and configuration sync between routing engines. The fabric link handles real-time session state between the packet forwarding engines. If you lose the fabric link the cluster degrades. If you lose the control link things get more complicated.
Step 1: Enable Cluster Mode
Chassis cluster is enabled from operational mode, not configuration mode. The command reboots the device immediately.
On SRX-A1:
root@SRX-A1> set chassis cluster cluster-id 1 node 0 rebootOn SRX-A2:
root@SRX-A2> set chassis cluster cluster-id 1 node 1 rebootBoth devices come back in cluster mode. The prompt changes to {primary:node0} on SRX-A1 and {secondary:node1} on SRX-A2. From this point, all configuration happens on Node 0 and changes sync to Node 1 automatically over the control link.
The cluster-id can be 1 through 255. Setting it to 0 disables cluster mode. The value matters because the shared virtual MAC address on reth interfaces is derived from it. Two clusters on the same broadcast domain must use different cluster-ids or you get MAC collisions.
Step 2: Configure the Fabric Link
The fabric link is the private channel between nodes for session state synchronization.
set interfaces fab0 fabric-options member-interfaces ge-0/0/1set interfaces fab1 fabric-options member-interfaces ge-7/0/1Node 1's interfaces appear as ge-7/0/x in the CLI. When two devices form a cluster they share one configuration, so Junos needs a way to distinguish Node 0's ge-0/0/0 from Node 1's ge-0/0/0. It solves this by mapping Node 1 to a different FPC slot (slot 7 on vSRX). The renaming happens automatically on reboot. fab0 takes Node 0's physical port, fab1 takes Node 1's.
Verify before moving on:
show chassis cluster interfacesExpected output:
Control link status: UpFabric link status: Up fab0 ge-0/0/1 Up / Up fab1 ge-7/0/1 Up / UpIf fabric shows Down, check the physical cable and confirm the correct interfaces are assigned.
Step 3: Set reth-count and Redundancy Group Priorities
Before any reth interface configuration can commit, you declare how many reth interfaces the cluster will use.
set chassis cluster reth-count 2Redundancy Group 0 manages the control plane and is created automatically. Redundancy Group 1 manages the data plane. This is where you set which node is preferred active. Higher priority wins.
set chassis cluster redundancy-group 1 node 0 priority 100set chassis cluster redundancy-group 1 node 1 priority 50Node 0 at 100 becomes primary. Node 1 at 50 stays standby. If Node 0 fails and later recovers, it will not automatically reclaim primary status unless you configure preempt. Without preempt, whichever node is currently primary stays primary until a manual failover or another failure event.
Step 4: Configure reth Interfaces
reth interfaces are logical interfaces that float between nodes on failover. Each reth has one physical child link per node. The IP address lives on the reth, not on the physical interface.
# Node 0set interfaces ge-0/0/0 gigether-options redundant-parent reth0set interfaces ge-0/0/2 gigether-options redundant-parent reth1# Node 1set interfaces ge-7/0/0 gigether-options redundant-parent reth0set interfaces ge-7/0/2 gigether-options redundant-parent reth1# Assign to redundancy groupset interfaces reth0 redundant-ether-options redundancy-group 1set interfaces reth1 redundant-ether-options redundancy-group 1# IPs on the rethset interfaces reth0 unit 0 family inet address 203.0.113.1/30set interfaces reth1 unit 0 family inet address 192.168.1.1/24One thing to watch in PNETLab: chassis cluster consumes two NICs before data interfaces appear. One for fxp0 (management) and one for em0 (control link). What was ge-0/0/0 in standalone mode becomes em0, and what was ge-0/0/1 becomes ge-0/0/0. With 4 NICs you end up with ge-0/0/0 and ge-0/0/1 as data ports. ge-0/0/2 requires a 5th NIC. The topology will show the connection but Junos will not have the interface until the NIC is actually allocated in the hypervisor.
Step 5: Restore IPsec Configuration
Everything from the previous IPsec lab transfers directly. The only change is that interface references update from physical to reth.
IKE gateway - update the external interface:
set security ike gateway GW-SRX-B external-interface reth0.0Security zones - update interface assignments:
set security zones security-zone UNTRUST interfaces reth0.0set security zones security-zone UNTRUST host-inbound-traffic system-services ikeset security zones security-zone TRUST interfaces reth1.0set security zones security-zone VPN interfaces st0.0IKE proposals, IPsec proposals, VPN objects, address books, policies, static routes are all identical to the standalone config. The st0 address stays the same. The tunnel peer address stays the same.
Full cluster config:
set chassis cluster reth-count 2set chassis cluster redundancy-group 1 node 0 priority 100set chassis cluster redundancy-group 1 node 1 priority 50set interfaces ge-0/0/0 gigether-options redundant-parent reth0set interfaces ge-0/0/2 gigether-options redundant-parent reth1set interfaces ge-7/0/0 gigether-options redundant-parent reth0set interfaces ge-7/0/2 gigether-options redundant-parent reth1set interfaces fab0 fabric-options member-interfaces ge-0/0/1set interfaces fab1 fabric-options member-interfaces ge-7/0/1set interfaces reth0 redundant-ether-options redundancy-group 1set interfaces reth0 unit 0 family inet address 203.0.113.1/30set interfaces reth1 redundant-ether-options redundancy-group 1set interfaces reth1 unit 0 family inet address 192.168.1.1/24set interfaces st0 unit 0 family inet address 10.255.1.1/30set security ike gateway GW-SRX-B external-interface reth0.0set security zones security-zone UNTRUST interfaces reth0.0set security zones security-zone TRUST interfaces reth1.0Step 6: Verify
Cluster health:
show chassis cluster statusIPsec tunnel:
show security ike security-associations # Phase 1show security ipsec security-associations # Phase 2ping 10.255.1.2 # st0 transitping 192.168.2.10 # end-to-end through tunnelManual failover:
# Fail over to node1request chassis cluster failover redundancy-group 1 node 1# Verify node1 is now primaryshow chassis cluster status# Reset manual mode lock, then fail backrequest chassis cluster failover reset redundancy-group 1request chassis cluster failover redundancy-group 1 node 0The reset is required. After any manual failover, Junos locks the redundancy group to prevent automatic bounce-back. Skipping it gives you Redundancy-group 1 is in manual failover mode already.
Thoughts
Honestly this was harder than I expected. The HA concept is not complicated. The actual configuration is not that bad either once the cluster is formed. What took the time was dealing with virtualization. NIC allocation, interface names shifting after cluster mode kicks in, a control link cable that silently disconnected when I edited the VM. None of that would come up on physical hardware. It is just the tax you pay for running virtual labs.
