How to Provide Redundancy with Junos Virtual Chassis

Because each switch member in an EX series Virtual Chassis unit has its own Routing Engine (RE), the Virtual Chassis unit has inherent redundancy. In addition, you can configure Graceful Routing Engine Switchover (GRES). Before explaining the difference between these two types of redundancy, you need to look at how the switch’s REs work.

In a Virtual Chassis unit, the master member acts as the master RE, running the routing protocols, providing the forwarding table that the Packet Forwarding Engines (PFEs, the lower layer processors) on all the member switches of the Virtual Chassis unit use to forward traffic on the LAN, and running management and control processes for the entire Virtual Chassis unit.

When you issue a commit synchronize command, the master RE sends the new configuration to the backup RE to ensure the configuration is synchronized; however, the backup RE does not actively run routing protocols or keep state with the master RE.

With redundant failover, when the master member fails, the backup RE assumes mastership and begins acting like the master RE (running routing protocols, building forwarding tables, and so on). Because the two REs haven’t been synchronized, this change is rather traumatic for the PFEs.

Imagine being in the middle of intensely reviewing a spreadsheet and suddenly having all the numbers change on you! For this reason, the PFEs on all the member switches in the Virtual Chassis reinitialize their state to the boot-up state before connecting to the new RE. After they reboot, everything is better, and they begin talking to the new RE.

GRES allows the transition to the new master RE to occur with minimal interruption in network traffic. When you configure GRES, the master and backup REs synchronize certain information. This synchronization allows the PFEs to seamlessly switch from one RE to another. The PFEs never reinitialize their state to the boot-up state, preventing a forwarding outage.

Configuring GRES requires a single command:

[edit]
user@junos-switch# set chassis redundancy graceful-switchover

Even though the switchover may be fairly seamless for the PFEs, the new master RE still needs to restart the sessions with all its routing protocol peers.

By default, the switchover will cause a forwarding outage while the old sessions are torn down and the new sessions are established. For this reason, on switches that perform Layer 3 routing, you’ll likely want to combine GRES with graceful restart, which allows Layer 3 forwarding to continue with the existing routing information while the new master RE starts sessions with routing protocol peers and builds routing and forwarding tables.

Once the new master RE has completed building new routing and forwarding tables, it sends updates to the PFEs. These updates prevent an outage while the new master RE gets up to speed. You configure graceful restart with this command:

[edit]
user@junos-switch# set routing-options graceful-restart

For graceful restart to work correctly, the routing protocol peers of the switch must support graceful restart in helper mode. Just like it sounds, devices that support graceful restart in helper mode will help peers that have failed and want to perform a graceful switchover.

These devices will maintain routes for a failed switch while the backup RE is taking over, resend all routing information to the new RE, and receive new routing information from the new RE. Once the restart is complete, the helper device will compare the routes it has received from the new RE with the routes it had received from the former master RE.

If it finds differences, only then will it update its routing table. Helper mode allows the network to keep forwarding traffic with the routing and forwarding tables that existed at the time of the switchover until the new RE has had a chance to build all its routing protocol adjacencies and exchange routes with them.

Thankfully, by default, all routers running the Junos OS software support graceful restart helper mode for all protocols except BGP. So, if you use only Juniper Networks routers and don’t run BGP, you don’t need to do anything else. You can configure the Junos OS software to support graceful restart for BGP simply by configuring graceful restart on the device.

If you use another vendor’s routers, you will need to check their documentation to determine how to enable their devices to run graceful restart for particular protocols.

blog comments powered by Disqus
Advertisement

Inside Dummies.com