Appendix E

Changing Clustered Hardware Nondisruptively

About this appendix

This appendix describes the process of changing hardware components using the nondisruptive cluster upgrade method.

For more information on nondisruptive upgrades of clustered systems, see Planning for upgrading clusters.

Note
See the Network Appliance hardware documentation for your filers before upgrading hardware components.

When hardware updates can be performed

Hardware maintenance and upgrades can be performed using the nondisruptive clustered upgrade method. Nondisruptive hardware updates can be performed independently or in conjunction with a Data ONTAP upgrade of clustered filers.

If hardware and software are updated together, the new software must be downloaded first. If you are upgrading within a release family, you can upgrade the hardware and Data ONTAP together. If you are upgrading between release families, you need to upgrade Data ONTAP first, which might cause a small disruption, but you can use the nondisruptive upgrade procedure to cover the longer period required for hardware upgrades.

In a nondisruptive upgrade, each node is successively "failed" and its partner is put in takeover state at one point in the procedure. While the "failed" node is in the waiting for giveback state, it can be interrupted and brought to the firmware ok prompt or powered off.

Supported hardware upgrades

When a cluster node is in "failed" mode, you can perform the following hardware upgrade procedures:

Replacing the head (with the same type as before, and with the same adapter cards)
Replacing the motherboard
Replacing or adding an adapter card
You can replace a nonvolatile RAM (NVRAM), disk, or network interface card (NIC) component, either with the same component or with an improved component (for example, you can upgrade from 2-port to 4-port Gigabit Ethernet (GbE), or 1-port to 2-port FC).
Replacing the Cluster Interconnect card
Installing onboard firmware on various platforms; for example, the Baseboard Management Controller (BMC) on 9xx filers.

You can also run diagnostics on various components of the "failed" machine (for example, motherboard or NVRAM).

Note
Running diagnostics on parts of the machine that are still visible to or in use by the partner--notably disks and the Cluster Interconnect card--can present problems. See the diagnostics documentation.

When these procedures are completed, you can reboot the node into the waiting for giveback state. In the rest of the nondisruptive upgrade procedure, the partner node will then do a cf giveback and the node will boot on its new hardware and firmware.