This document describes the process to replace the chassis of a Nexus 9500 spine switch, which failed due to a hardware issue, in Application Centric Infrastructure (ACI) mode.
There is a current and working ACI fabric and a Nexus 9500 spine has failed. The failure has been determined to be a hardware failure and only the chassis needs to be replaced. All working components (supervisors, line cards, fabric modules, and so on) need to be transferred into the replacement.
Complete these steps in order to replace the spine:
- Disconnect all power cables in order to ensure the failed spine is powered off.
- From the Application Policy Infrastructure Controller (APIC) GUI, choose Fabric > Inventory > Unreachable Nodes. The spine is listed here after it has been powered off for a few minutes. Take note of its 'Name' and 'Node ID'.
- Decommission the failed spine and remove it from the controller.
- Select the spine in the work pane. From the Actions drop-down list, choose Decommission.
- Click the Remove from controller radio button and click OK.
- Remove the failed spine from the rack and install the replacement chassis. Transfer all working components into the new chassis in the same arrangement as in the failed chassis. Components to transfer include fabric modules, supervisors, system controllers, power supplies, fan trays, and line cards.
- Console into the active supervisor in the new chassis in order to erase its configuration.
- Enter the cat /proc/cmdline command in order to determine which ACI image currently runs on the supervisor. Look for the 'ksimg' section.
- Use the ACI switch image name and enter these commands:
spine# /bin/prepare-mfg.sh <aci_image.bin>
- If the reload window was missed and you are unable to reload the device, enter the vsh -c reload command.
This erases the configuration on the active supervisor. Repeat Step 5 in order to erase the configuration on the standby supervisor. The configuration is automatically pushed down from the APIC after the spine joins the fabric.
- In the APIC GUI, choose Fabric > Inventory > Fabric Membership. The spine is listed as a row in the work pane without a Node ID, Node Name, and IP address. Double-click the row and assign the same 'Node ID' and 'Node Name' as before.
The spine receives an IP address within a few minutes, which is reflected in the APIC GUI.
The spine has fully rejoined the ACI fabric and the APIC automatically pushes down all relevant policies. In this case, the chassis of a spine with Node ID 201 failed. The chassis was replaced and its serial number was mapped to the same Node ID. The APIC then pushes all policies relevant to Node 201 to the spine without the need for further configuration.