This section explains how to safely replace hardware components on Cisco platforms running Qumulo Core.

Component Replacement Overview

Component Hot-Swappable Node Offline Required
Drives (NVMe, SSD, HDD) Yes No
Boot Drive No Yes (special procedure)
Power Supplies Yes No
Fans Yes No (but special handling required)
DIMMs (Memory) No Yes
CPUs No Yes
Motherboard No Yes
NICs No Yes

Hot-Swappable Components

Drive Replacement

For the drive replacement procedure, see Drive Replacement.

Power Supply Replacement

Power supplies are hot-swappable when the server has redundant power supplies.

  1. Verify the server has redundant power supplies and the remaining PSU is functioning.
  2. Remove the failed power supply.
  3. Insert the replacement power supply.

Fan Replacement

Fans are hot-swappable but require prompt replacement to prevent thermal issues.

  1. Identify the failed fan using CIMC or front panel LEDs.
  2. Have the replacement fan ready before beginning.
  3. Follow the fan replacement procedure in the Cisco Installation and Service Guide for your server model. This includes removing the top cover and accessing the fan modules.

For detailed fan replacement procedures, refer to the Cisco Installation and Service Guides listed in Platform Information.

Components Requiring Node Offline

For DIMM, CPU, motherboard, or NIC replacements, the node must be safely taken offline using the following procedure.

Before You Begin

Step 1: Verify Cluster Protection Status

You can verify cluster protection status using either the Web UI or the CLI.

Using the Qumulo Web UI

  1. Log in to the Qumulo Web UI.
  2. Navigate to Cluster > Cluster Overview.
  3. Under Data Protection, verify that the cluster can tolerate a node failure.

Using the qq CLI

SSH into any node and run:

qq protection_status_get

You will see output similar to:

{
    "remaining_drive_failures": 1,
    "remaining_node_failures": 2
}

Step 2: Recuse the Node

Recusing a node safely removes it from the cluster quorum, allowing data protection mechanisms to account for the missing node.

  1. SSH into the node that requires maintenance.

  2. Run the recuse command:

    /opt/qumulo/recuse_node.py --reason "Component replacement"

  3. After running this command, a red banner appears in the Qumulo Web UI stating:

    Unable to communicate with node X

  4. This confirms the node was successfully recused from the cluster.

Step 3: Power Off the Node

After the node is recused, power it off:

Using CIMC

  1. Log in to the node’s CIMC interface.
  2. Navigate to Server > Power.
  3. Click Power Off.

Using the CLI

From another node or management system:

ipmitool -I lanplus -H <CIMC-IP> -U <your_user> -P <password> power off

Using the Physical Power Button

Press and hold the power button on the front panel until the server powers off.

Step 4: Replace the Component

Perform the hardware replacement according to your hardware vendor’s documentation:

DIMM Replacement Notes

When replacing DIMMs:

  1. Follow your vendor’s DIMM population guidelines.
  2. Ensure replacement DIMMs match the specifications of the original.
  3. Use proper ESD precautions.

Step 5: Power On the Node

After completing the component replacement:

  1. Power on the node using the power button, CIMC, or IPMI command.
  2. Wait for the node to complete POST and boot Qumulo Core.
  3. Verify the component is functioning:
    • For DIMMs: Check total memory in CIMC or using qq node_status
    • For NICs: Check network connectivity
    • For fans: Check CIMC sensor readings

Step 6: Reintroduce the Node

After verifying the component replacement was successful, reintroduce the node to the cluster:

  1. SSH into the node that was serviced.

  2. Run the reintroduce command:

    /opt/qumulo/sbin/reintroduce_node.sh

  3. Monitor the Qumulo Web UI to verify:

    • The red banner disappears
    • The node appears healthy in the cluster overview
    • Data reprotection begins (if applicable)

Step 7: Verify Cluster Health

After reintroduction:

  1. Check cluster protection status:

    qq protection_status_get

  2. Verify all nodes are healthy:

    qq cluster_slots

  3. Monitor the Web UI for any alerts or warnings.

For Boot Drive replacement (a special procedure), see Boot Drive Replacement.

Getting Help

If you have any questions or encounter issues during component replacement, contact contact the Qumulo Care Team through Slack, email, or by phone.

For Cisco hardware-specific issues, contact Cisco Support.