Replacing a unit in a Securestack:
The first consideration is firmware.
Load the appropriate firmware on the new unit before adding it into the stack. If the firmware on production and replacement are on opposite sides of the 6.42/6.61 line, you *must* load the appropriate code onto the replacement in stand-alone first. Unfortunately, 6.42 and 6.61 do not communicate over the stacking ports.
Pre-provision the replacement with the appropriate firmware, and the two firmwares in question are on the same side of 6.42/6.61, then the code should automatically be pushed from the manager to the new member. If this process does not happen automatically the following command can be used:
set switch copy-fw destination-system [unit #]
Here are the simple steps to replace a member unit:
The steps to replacing a manager unit :
- Pull power from the member that is being replaced.
- Pull stacking cables from the member that is being replaced.
- Label Ethernet cables if not already labeled.
- Pull Ethernet cables from the member that is being replaced.
- Remove member from rack.
- Place replacement member into the rack.
- Plug Ethernet cables into replacement member.
- Plug stacking cables into replacement member.
- Plug power into replacement member.
- Wait for Ethernet port LEDs to light up. If they do, you are done. Check connectivity to network resources. If they do not, move to next step.
- Check "show switch" in the CLI of the manager switch. If there is a code mismatch, issue the "set switch copy-fw ..." command that was given above.
- After your CLI prompt returns reset the individual member. "reset [unit #]"
- When the switches finish booting, Ethernet LEDs should light up. Check connectivity to network resources.
Make sure you have a backup of the configuration saved to TFTP server: How to backup configuration from old Securestack switch to replacement Securestack switch
To swap out the operational manager unit for a replacement manager unit having the same unit number and firmware version, these are the steps that would be required:
- Move management from the faulty manager to the working member (30-40 seconds stack disruption)
"set switch movemanagement [from-unit] [to-unit]"
- Remove the now non-manager unit (2-3 seconds disruption)
- Add a new non-manager unit (will eventually integrate - takes longer if the stack is busy), and confirm that it has the proper unit number.
- Issue the 'set switch movemanagement...' command to swap the manager role to the unit that was just added (30-40 seconds stack disruption).
In none of these steps would an actual reset (power cycle) of stack units occur. However, the entire stack will not be able to pass traffic for certain intervals of time as stated above in steps 1, 2, and 4.