Rack Installation

This guide covers connecting an Oxide rack to your network after physical installation and power connection are complete. For physical installation procedures (uncrating, positioning, power connection), see the Facilities Guide.

Tip
On installation day, facilities teams complete physical installation and power connection first. After power validation, the network team follows this guide to establish network connectivity.

Prerequisites

Confirm that the following facilities work is complete before proceeding:

And that all rack LEDs are solid:

  • All power shelf LEDs are solid (no blinking or red/amber indicators).

  • All sled power LEDs are solid.

  • Both switches power LEDs are solid.

Additionally, confirm that the following are on-site and ready:

  • Requested compatible (transceivers) are present and plugged into the rack.

  • Network uplink fiber and RJ45 cables for technician port(s).

  • Jumpbox or technician laptop with IPv6 SLAAC enabled.

Note
If power validation failed or any LEDs show abnormal behavior, return to Facilities: Power Connection to troubleshoot before proceeding with network connectivity.

Connect Uplink Fiber

If not already present, insert the optical transceivers into the planned QSFP ports on each switch and connect the fiber patch cables from your upstream network equipment to the patch panel on the top of the rack.

Depending on your rack version, the patch panel layout may differ. Refer to the appropriate diagram for your rack:

Oxide Rack v1
Oxide Rack (Version 1)
Oxide Rack v2
Oxide Rack (Version 2)

Verify the following:

  • The switch and QSFP port numbers you are using match what you will configure in the rack.toml (e.g., if you plug into port 0 on switch 0, you configure qsfp0 for switch 0).

  • The fiber connectors are clean — use a fiber cleaning tool if available.

Note
Pay close attention to the schematic guide for the fiber patch panel to ensure you are connecting to the correct ports. The patch panel is labeled with port numbers that correspond to the switch QSFP ports, but the physical layout may be non-intuitive. See the diagram for your rack version to confirm the mapping between physical connector positions and logical QSFP port numbers.

Check Transceiver LEDs

Once the uplink connections have been established, verify that the LEDs are lit for the plugged-in transceivers.

An off or blinking LED indicates a physical connectivity issue. Use the following to diagnose:

SymptomCauseSolution

LED completely off.

Unsupported transceiver, not fully seated, or faulty.

Verify transceiver is on the supported list, reseat the transceiver, try a known-good replacement.

LED blinking.

Intermittent physical connection.

Clean and reseat the fiber, check for bent pins or damaged ferrules.

LED solid but no link on upstream.

Speed/FEC mismatch with upstream port.

Verify that the upstream port speed and FEC settings match what you will configure (e.g., 100G with RS FEC).

Tip
If you’re following along in sequential order, transceiver signal quality (Rx/Tx power levels) is verified in detail during the Initial Rack Setup phase using the Wicket TUI. At this stage, you are only confirming whether the LEDs indicate basic physical connectivity.

Connect Technician Ports

Connect RJ45 cables from your jumpbox or technician laptop to the technician ports on the rack switches. Each switch has two technician ports (labeled T0 and T1) on its front panel. They are on the right side of the switch, and are the only RJ45 ports on the front.

Tech Ports
Note
It isn’t required for the technician ports to be connected in order to operate the rack, but you will need at least one technician port connection per switch in order to perform the initial rack setup and configuration via the Wicket TUI. This is because the technician ports provide out-of-band management access directly to each switch (and the entire rack).

Technician Port Isolation

Important
The two technician port connections — one on either switch — must be isolated from each other at the network level, meaning that packets exiting the technician port on one switch must not be able to reach the technician port on the other switch.

If both tech port cables are connected to the same physical upstream switch, then that upstream switch must be configured to place them on separate VLANs.

Why isolation matters: If tech ports are not isolated, both switches will advertise IPv6 addresses on the same broadcast domain. This causes:

  • The Wicket TUI to only discover one switch instead of both.

  • Race conditions during initialization where the Management Gateway Service cannot reliably determine which switch it’s communicating with.

  • IPv6 neighbor discovery conflicts that can cause connection timeouts.

  • Failed rack initialization with errors like "cannot reach scrimlet on switch1" or "MGS connection refused".

When properly isolated, each tech port operates in its own network segment, allowing wicket to discover and communicate with both switches independently.

Verify IPv6 Connectivity

The technician ports send out periodic IPv6 SLAAC advertisements. To verify connectivity from the jumpbox or laptop:

  1. Ensure the connected interface is configured for IPv6 autoconfiguration (SLAAC).

  2. Check for received SLAAC advertisements:

    ip neighbor show | grep fdb1

    You should see output like:

    fdb1:a840:2504:195::1 dev eno2 lladdr 02:08:20:36:5c:8d REACHABLE
fdb1:a840:2504:352::1 dev eno1 lladdr 02:08:20:bb:26:4d REACHABLE

  3. Verify SSH connectivity to the wicket shell:

    ssh wicket@fdb1:a840:2504:195::1

    You should then see an Oxide splash screen. Press Ctrl-C to exit for now.

If you do not see any fdb1: addresses:

SymptomCauseSolution

No fdb1: addresses appear.

IPv6 SLAAC not enabled on the interface.

First verify IPv6 is enabled: ip -6 addr show <interface> should show an inet6 address. If no IPv6 addresses appear at all, enable IPv6 autoconfig on the jumpbox NIC (e.g., on Linux: sysctl -w net.ipv6.conf.<interface>.accept_ra=1).

Addresses appear but SSH times out.

Firewall blocking TCP port 22.

Check firewall rules between jumpbox and tech port network.

Addresses appear but SSH connection refused.

Switches still booting.

Wait 2–3 minutes and retry.

Only one address (not two).

One switch cable not connected or tech port isolation misconfigured.

Check the cable to the other switch; verify VLAN isolation is correct.

Tip
For a visual troubleshooting workflow, see the decision tree diagram below:
image

macro when ready.

Troubleshooting: Can't SSH to Tech Port
Troubleshooting: Can’t SSH to Tech Port

Once you have confirmed SSH connectivity to wicket on at least one tech port per switch, the physical installation is complete.

What’s Next

Network connectivity is established. Proceed to:

  • Phase 3: Initial Rack Setup — Component validation, software updates, rack.toml configuration, and rack initialization via wicket (2-4 hours with a network engineer on standby).

Have your Phase 1 Summary Checklist ready — you’ll need the DNS servers, NTP servers, IP ranges, and routing configuration you prepared in Phase 1.