Description

This article provides information about the behavior of the Virtual Chassis, when a split occurs. 

Symptoms

Information about the behavior of the Virtual Chassis, when a split occurs.

Solution


Virtual Chassis is a concept, in which multiple switches are stacked together to make them appear as a single switch. Up to 10 EX4200 switches can be interconnected via dedicated Virtual Chassis ports on each device or optional uplink module ports that are configured as Virtual Chassis ports, with a combined backplane bandwidth of up to 128 Gbps.

Solutions that use the EX4200 line with Virtual Chassis technology combine the scalability and compact form factor of stand-alone switches with the high availability, high backplane bandwidth characteristics and high port densities of traditional chassis-based switches.

For resiliency and redundancy, the Virtual Chassis configuration includes a primary and a backup switch; both dynamically elected as part of the Virtual Chassis configuration. Each remaining switch serves as a line card switch; but is ready to be selected as a backup switch, if the primary or backup switch fails.

Outages when upgrading a Virtual Chassis :

There will be outage, because as in a Virtual chassis, multiple switches are stacked and made to work as a single switch; so the code must be the same on all of the devices. It will not be possible to upgrade a particular member and then upgrade the rest; so outage occurs, as the switch needs to be rebooted for the new code to be activated on it.


When Primary switch fails : 
  ------------------------------------------------------
  |                                                    |
  |                   Primary                           | 
  ......................................................
                   |             |
  -------------------------------------------------------
  |                                                   |.................... PC-A
  |                  BACKUP                           | 
  ........................................................
                 |              |
  ------------------------------------------------------
  |                                                     |...................PC-B
  |                 LINE CARD                           | 
  ........... ..........................................
In this scenario, all routes and the forwarding state information are maintained by the Primary routing engine on enabling GRES. If the Primary switch reboots, then the Backup becomes the next Primary switch and it presumes the forwarding traffic; but that takes some down time. The range is from 4-7 RTO, if a continuous ping is performed from PC-A to PC-B .

The reason is that as the primary goes down, the backup switch must become the primary and the information must be synchronized. Similarly, a new Backup must be elected.

When the Line Card fails :
  ------------------------------------------
|                                         |...........PC-A
|           Primary                        | 
...........................................
        |                 |
--------------------------------------------
|                                          |..........PC-B
|           BACKUP                         | 
............................................
       |                  |
-------------------------------------------
       |                  |
       |    LINE CARD     | 
............................................
In this scenario, nothing happens, as the kernel state information and the forwarding state information are already present in the Primary routing engine and the PC is connected to the Primary and the Backup switch.

Here, the linecard is disconnected and this does not cause any interruption for the continuous ping from PC A to PC B .

When the backup is disconnected :
-----------------------------------------
|                                       |................PC-A
|             Primary                    | 
.........................................
           |             |
------------------------------------------
|                                        |
|             BACKUP                     | 
...........................................
           |             |
-------------------------------------------
|                                         |..............PC-B
|             LINE CARD                   | 
...........................................
In this scenario, the backup is connected to the Line card and the Primary switch and the backup are rebooted. Also, the Primary goes into a linecard’s position that detects a VC split.

The reason why the Primary routing engine takes up the role of a Linecard is to avoid the possibility of dual Primary formation in the topology, in the event of a VC split.

Complete outage will occur.

For this topology : 
|.....................................|
|   ----------------------------------| 
|   |                                 |............... PC-A
|   |         Primary                  | 
|   ...................................
|      |                                         
|   -----------------------------------
|   |                                 |
|   |         BACKUP                  | 
|   ...................................
|                               |
|   ----------------------------------
|   |                                |..............PC-B
|   |         LINE CARD              | 
|   .................................. 
|......................|

For the topology, as depicted above, the VCP cables are connected in this format and two PCs are connected to the Primary and the linecard and if the backup switch reboots, then again, complete network outage will occur; as the Primary detects the VC split condition and it will take up the position of a linecard.

But, if split detection is disabled and even though the Backup switch is rebooted, the Primary stills remain the Primary and the packet flows through; without any drops.

The packet flow in a VC is completely dependent on the lowest cost path to the destination. if a ping is performed from PC-A to PC-B , the packet has two routes to reach the destination; one is via the Backup switch and the other is directly to the linecard. It will prefer the least cost path, which is the link that connects the linecard and the Primary together.

So, if the Backup switch is rebooted, the packet will still continue to flow through the same link and packet loss will not occur. The question regarding physical disconnect of the LAN cable is not possible, as the link will remain intact; as long as the unit, to which the PC is connected, is not rebooted.

The deployment of the switches in the stack requires a proper understanding of the above information, so that the minimum interruption of traffic might occur.