Updated: 20 November 2026
StackWise is mostly plug-and-play, the cabling does most of the work, but there are a few configuration steps that decide whether the stack forms cleanly, which switch leads, and how the members are numbered. Here's the whole process, in order, with the commands.
Three conditions must be true or the stack won't form. Every switch must be the same model family, run the same IOS XE software version, and hold the same licence level (Essentials with Essentials, Advantage with Advantage). You cannot mix a 9200 with a 9200L, or a 9300 with a 9300L; a 9300 and 9300X can stack, but the whole ring runs at the slower speed.
Check the hardware too. The Catalyst 9300 has built-in rear stacking ports and uses StackWise cables directly. The 9300L needs an external stack kit (a network module and stacking cables) rather than built-in ports. Have the right cables for your model before you begin.
Power the switches off, then connect the StackWise cables in a ring, not a chain. Connect stack port 1 on each switch to stack port 2 on the next, and then close the loop by connecting port 1 on the last switch back to port 2 on the first.
Closing the ring matters. If you leave it as a half-ring (a daisy chain), you lose 50 per cent of the stack bandwidth and all path redundancy, one cable failure would then split the stack. Always close the ring.
Power the switches on. StackWise forms the stack automatically and elects an active switch and a standby based on priority and other tiebreakers, so you don't configure the stack into existence, the cabling and boot do that. Give it a couple of minutes to settle.
Confirm every member joined and the ring is healthy:
Switch# show switch
Switch# show switch stack-ports summary
Switch# show switch neighbors
show switch lists each member with its number, role (Active, Standby or Member), priority and state, you want every switch showing Ready. show switch stack-ports summary confirms both stack ports on each member are up, which tells you the ring is complete rather than half-open.
By default every switch has priority 1, and you want to decide which switch leads rather than leave it to chance. Set the intended active switch to the highest priority and work down. Priority is set from the active switch in EXEC mode:
Switch# switch 1 priority 15
Switch# switch 2 priority 14
Switch# switch 3 priority 13
The valid range is 1 to 15, and higher wins. Note that a priority change affects the next election, it doesn't force an immediate switchover, so the active role moves at the next reload rather than instantly.
If the physical order and the switch numbers don't match what you want, renumber a member. This also runs from the active switch in EXEC mode and requires a reload of that member to take effect:
Switch# switch 3 renumber 2
Keep the numbering logical (top-of-rack to bottom, say), so the stack is easy to reason about later.
You can configure a switch's ports before it physically joins the stack, useful when you're staging configuration ahead of adding hardware. Provisioning is done in global configuration mode with the member number and model:
Switch(config)# switch 4 provision c9300-48p
When switch 4 is later cabled in, it inherits the configuration you prepared.
Save the configuration so the priorities, numbering and provisioning survive a reload:
Switch# copy running-config startup-config
Nine times out of ten it's one of the rules from the start. Confirm the member is running the exact same IOS XE version and licence level as the rest, check show switch stack-ports summary to confirm both stack cables are seated and the ring is closed, and confirm you're not trying to mix incompatible models (9200 with 9200L, or 9300 with 9300L). Align the version and licence, reseat the cables, and the member joins.
On the Catalyst 9300, StackPower is separate from StackWise data stacking. It pools the power supplies across the stack over dedicated StackPower cables, so a member that loses its own power feed can draw from a neighbour, valuable when the stack runs dense, high-wattage PoE. It's cabled and configured independently of the data stack, so treat it as its own step if you need power redundancy.
We'll get back to you shortly.