Appendix B | About Switch Stacking |
It should be emphasized that when two stacks are combined, all of the configuration information for one of the stacks will be lost. Only the surviving master (after the discovery/election process completes) will maintain its configuration information.
The best practice to combine two stacks is to reset the switches in one stack to the factory defaults and then add the switches as described in the “Adding Units to a Running Stack” subsection of section “Normal (Self- Ordering) Stack).”
•If one of the merged stacks had neither a Master unit nor a Backup Master unit, then units belonging to this group will be inserted into the stack in the exact way as described in section “Replacing a Failed Stack Member in a Running Stack” above. The Master will either connect the running units to the stack using the current numbers or will renumber them as necessary. The process described in section ““Replacing a Failed Stack Member in a Running Stack” applies to this case as well.
It should be emphasized that any time two stacks are combined into one stack, there is no way to maintain the configuration for both sets of switches. All dynamic information of the units that belong to the portion of the stack that was not reelected to be the master will be relearned.
Stacking Cable Failure
In this example, let us assume that stacking connection cables failed and caused a stack split, as described in section “Splitting a Stack.” When the stacking cable connection is fixed and units are reconnected, it results in merging two stacks as described in section “Merging Two Stacks.”
This scenario is feasible only if the topology of the stack is Chain topology. Single stacking cable failure will not cause a stack split if a Ring topology is used.
Inserting Too Many Units
In this example, a user tries to insert too many units into a stack.
1.All units (existing and newly inserted) are powered on at the same time:
•A Master is elected following the Master Discovery and Master Election processes.
•All other units will shut down.
NOTE: In some extreme cases, due to a race condition during the boot process, some of the units might be connected and join the stack.
2.A running group of units is added to an existing stack, assuming each one of the stack groups has an elected Master. The total of existing units and inserted units would exceed the maximum allowed number of units in a stack, which is 6 units for SLM224G4PS, or 4 units for SLM248G4PS:
•Master Detection and Master Election processes would determine the master out of one of two combined stacking groups.
•When switches are added to a running stack, the Unit ID Allocation and Duplicate ID Conflict Resolution process will detect an error if too many switches are present in the stack, and no changes will be made to units that originally belonged
to the group managed by the newly elected master. The original switches will retain their ID assignments and configurations. The units that originally belonged to the group managed by the master that lost its master status will be shut down.
Standalone Unit Inserted into a Running Stack
Since the unit is in standalone mode it will not participate in a master discovery process (it will not look for a master and will not respond to master queries). As a result it will not join the stack but will continue to run as a standalone manageable unit.
The ports that are connected to the other units’ stacking links will not pass any traffic, and the master will consider them as failed stacking links and route all traffic around them.
58 |