The Master Switch; Splitting And Merging One Stack - IBM RackSwitch G8000 Application Manual

The Master Switch

Splitting and Merging One Stack

150
RackSwitch G8000: Application Guide
• Backup
One member switch can be designated as a Backup to the Master. The Backup
takes over control of the stack if the Master fails. Configuration information and
run-time data are synchronized with the Master.
An operational stack can have only one active Master at any given time. In a normal
stack configuration, one switch is configured as a Master and all others are
configured as Members.
When adding new switches to an existing stack, the administrator must explicitly
configure each new switch for its intended role as a Master (only when replacing a
previous Master) or as a Member. All stack configuration procedures in this chapter
depict proper role specification.
However, although uncommon, there are scenarios in which a stack may
temporarily have more than one Master switch. If this occurs, one Master switch will
automatically be chosen as the active Master for the entire stack. The selection
process is designed to promote stable, predictable stack operation and minimize
stack reboots and other disruptions.
If stack links or Member switches fail, any Members which cannot access either the
Master or Backup are considered isolated and will not process network traffic (see
"No Backup" on page 152). Members which have access to a Master or Backup (or
both), despite other link or Member failures, will continue to operate as part of their
active stack.
If multiple stack links or stack Member switches fail, thereby separating the Master
and Backup into separate sub-stacks, the Backup automatically becomes an active
Master for the partial stack in which it resides. Later, if the topology failures are
corrected, the partial stacks will merge, and the two active Masters will come into
contact.
In this scenario, if both the (original) Master and the Backup (acting as Master) are
in operation when the merger occurs, the original Master will reassert its role as
active Master for the entire stack. If any configuration elements were changed and
applied on the Backup during the time it acted as Master (and forwarded to its
connected Members), the Backup and its affected Members will reboot and will be
reconfigured by the returning Master before resuming their regular roles.
However, if the original Master switch is disrupted (powered down or in the process
of rebooting) when it is reconnected with the active stack, the Backup (acting as
Master) will retain its acting Master status to avoid disruption to the functioning
stack. The deferring Master will temporarily assume a role as Backup.
If both the Master and Backup are rebooted, the switches will assume their originally
configured roles.
If, while the stack is still split, the Backup (acting as Master) is explicitly reconfigured
to become a regular Master, then when the split stacks are finally merged, the
Master with the lowest MAC address will become the new active Master for the
entire stack.