User guide

Virtual Chassis Overview Configuring Virtual Chassis
page 12-12 OmniSwitch AOS Release 7 Switch Management Guide March 2015
• If the primary CMM on the Master chassis fails the secondary CMM, if available, will takeover and the
chassis will remain the Master chassis.
• If all CMMs on the Master chassis fail the chassis will reboot and the first-in-line Slave chassis will
take over becoming the new Master chassis. The first-in-line is derived from the same election criteria
that were used to select the original Master.
Split Chassis Detection
Split chassis detection is implemented using a proprietary protocol called RCD (Remote Chassis Detec-
tion) protocol. The goal of the split-chassis detection mechanism is to provide information in a virtual
chassis environment which can be used to determine whether a VFL has failed. A split chassis can occur
when the VFL connection is broken but each of the switches remains operational. This scenario must be
detected so that one of the switches remains the Master and continues using the same IP and MAC address
in the network.
Note. RCD is only enabled once the virtual chassis is operational. If a switch is unable to join a virtual for
any reason, the RCD protocol will not be enabled.
To help detect this scenario each switch in the Virtual Chassis topology periodically sends information via
its local EMP port. All of the switches participating in a Virtual Chassis should be able to communicate
via the local EMP port using an out-of-band network. When a VFL goes down, each switch can still
communicate with the others via the EMP port, this acts as a backup mechanism to help detect the split
chassis scenario. RCD will use the following IP addresses in order of preference:
1 CMM IP address stored in NVRAM (if configured)
2 Chassis EMP IP address
See the “Configuring EMP IP Addresses” on page 12-22 for information on configuring the EMP IP
addresses. Also, see the “Split Chassis Detection - OS10K CMMs” on page 12-13 for information on EMP
communication between CMMs.
Having both switches with all the front-panel ports up while the VFL is down can cause layer 2 and layer
3 connectivity issues. In order to avoid this scenario an out-of-band management protocol has been imple-
mented on the EMP port that detects the status of each chassis. If all VFL links go down then this protocol
will detect and shutdown all user ports on the former Slave chassis to prevent the duplicate IP and MAC
addresses from being used on the network. The user ports will automatically come up when the VFL
connectivity is re-established.
EMP
Out-of-band
Management
Each chassis in the Virtual Chassis
sends periodic updates via the EMP
port. Each chassis should be able to
communicate with the other via the EMP
out-of-band network.