57
About Switch Stacking
24/48-Port 10/100 + 4-Port Gigabit Smart Switch with Resilient Clustering Technology and PoE
Appendix B
before the split. Since the Backup Master was not
acting as a master prior to the split, it will initiate
a topology database and port-learning process.
Traffic might be halted for a short period of time
until synchronization (unit and port configuration)
is completed. New units learned by the Backup
Master will notify the system administrator (using
SYSLOG messages and SNMP traps).
In either case (steps b and c above), the subgroup
will continue to work as it did before the split,
except that the number of units is lower than prior
to the split.
No unit ID changes are made in either subgroup.
Each new stack will have a Master (one has the
original Master, the other has the Backup). They
thus operate as two separate stacks, both having
the same configuration and hence the same IP
address.
WARNING:
Both resulting stacks will have the
same IP Address. This may lead to problems on
the network, since there would be no way for
users to connect to one of the stacks through
its IP address.
Subgroup Contains Neither Master Unit nor Backup Master
Unit
Please note that this is exactly the same as the case of a
failed master, where no backup is available.
In this case the units with ID 3-6 (3-4 for SLM248G4PS)
in this subgroup will not renumber themselves, and
will remain shut down until a stack Master enabled unit
is connected, and starts to operate as stack master. It
is the responsibility of the Master-discovery process to
see that the master is gone.
In this subgroup, the units lose connection with the
Master. Since they started as a running stack and none
of them are in factory default mode, renumbering will
not take place, and even a Reset of the units will not
affect unit ID assignment (As noted above, units can
be renumbered only by a stack Master).
No unit ID changes are made in either subgroup.
Again it should be emphasized that none of the units in
either half of the stack will renumber themselves.
Merging Two Stacks
In this example, the user would like to merge two working
stacks and create one stack out of those two stacks. In this
example one should distinguish between two scenarios:
d.
e.
f.
•
•
•
The incoming units are turned off before insertion and
then powered back up afterward
The two stacks are kept running during the insertion
(for example, the stacking cables of the two stacks are
connected)
Incoming Units Turned Off Before Insertion, then Turned On
After Insertion
This is exactly the same as inserting units into a running
stack (refer to “Replacing a Failed Stack Member in a
Running Stack”). The only difference is that more than
one unit will be inserted into the stack and therefore for
each unit inserted to the stack the same process will be
executed.
Both Stacks Kept Running During Insertion
If each of the joined stacks has a Master unit, both Master
units will perform Master discovery and participate in a
Master Election, as described above. One of the Master
units will be selected as the Master unit—for the merged
stack. The criteria for choosing the master are:
Force Master
System Up Time
Lowest Unit ID
Lowest MAC
The process of master selection between the two master
units is as follows:
If force master is enabled then the unit which is forced
is chosen.
Up time is measured by quantities of periods of 10
minutes. If the number of 10 minute periods is higher
for one of the units then this unit is chosen.
If both units have the same up time (measured in
periods of 10 minutes) the unit with the lowest unit
ID is elected.
If both Master unit IDs are equal the unit with the
Lower MAC is chosen.
The Master unit that loses its “mastership” in the Master
election process will be renumbered if the unit ID was
dynamically allocated. renumbering will now allocate
it a new number by the new Master, as a stack member,
or possibly Backup Master. It should be emphasized
that in no case will there be two units with the same
Unit ID at the end of this process.
The Master unit that loses its mastership in the Master
election process will be shut down if the unit ID was
manually allocated. It is recommended that the
administrator configure it to auto-assigned Unit ID
before reconnecting it to the stack.
•
•
•
•
•
•
•
•
•
•
•
•
