Instance creation fails in dogfood console

[sion42x]

Instance creation in oxide silo throws on-screen error: Not found: default unicast IPv6 pool for current silo

Also mentioned in: oxidecomputer/omicron#9683 (comment) and documented here per @ahl

Not sure if this is an in-flight issue that will be fixed in the next update or not, but documenting it here for visibility.

I attempted a workaround to create the instance with no network interface (successful). However, when I continued to add the network interface to the stopped instance, the error was thrown:

The network interface with ID '26e58cc4-1316-44cd-a12e-0e879bc82cf3' is a dual-stack NIC, with IPv4 address '172.30.0.5' and IPv6 address 'fd2a:c6e9:d592::5'. However, the version of the client being used is unable to fully represent both IPv4 and IPv6 addresses. Update your client and retry the request.

[linear]

PRD-247

[ahl]

Maybe dup of oxidecomputer/omicron#9683

[sudomateo]

PRD-247

I disabled this feature via Linear, FYI. Apologies for the noise.

[sion42x]

The instance created via workaround (no network interface, failed addition after) now cannot be viewed:

CLI shows nothing odd:

stevenkaram@MacBook-Pro-2 oxide.rs % oxide instance view --instance 2c138c2d-8d40-49b7-9a86-ed76f58b362f --profile oxide2
{
  "auto_restart_enabled": true,
  "boot_disk_id": "1acb63cb-8274-4712-99bf-b7f1afa651f1",
  "description": "",
  "hostname": "fdasfdasfa",
  "id": "2c138c2d-8d40-49b7-9a86-ed76f58b362f",
  "memory": 8589934592,
  "name": "fdasfdasfa",
  "ncpus": 2,
  "project_id": "9a15ea69-d94b-4c04-9780-d649f4636f55",
  "run_state": "stopped",
  "time_created": "2026-01-20T17:58:25.446765Z",
  "time_modified": "2026-01-20T17:58:46.796841Z",
  "time_run_state_updated": "2026-01-20T17:58:52.326157Z"
}

Subsequent creation of an instance with no network interface is fine, can access the Instance in console without issue. However, attempting to add a network interface (fails with the same error in the original post) affects the instance the same way, making it impossible to view in console.

The nexus error after failing a network interface addition is the same as the error when trying to create it:

02:29:17.907Z INFO 444c69ba-2003-4ae0-a5dd-a6ba5d3a5fdf (dropshot_external): request completed
    error_message_external = The network interface with ID '721f9da1-1d04-4809-89c0-c87520b39af1' is a dual-stack NIC, with IPv4 address '172.30.0.6' and IPv6 address 'fd2a:c6e9:d592::6'. However, the version of the client being used is unable to fully represent both IPv4 and IPv6 addresses. Update your client and retry the request.
    error_message_internal = The network interface with ID '721f9da1-1d04-4809-89c0-c87520b39af1' is a dual-stack NIC, with IPv4 address '172.30.0.6' and IPv6 address 'fd2a:c6e9:d592::6'. However, the version of the client being used is unable to fully represent both IPv4 and IPv6 addresses. Update your client and retry the request.
    file = /home/build/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/dropshot-0.16.6/src/server.rs:855
    latency_us = 303871
    local_addr = 172.30.2.6:443
    method = GET
    remote_addr = 172.20.16.150:52773
    req_id = 8aee96a6-2856-4d9a-904f-258306847387
    response_code = 400
    uri = https://oxide.sys.rack2.eng.oxide.computer/v1/network-interfaces?project=skaram&instance=fdsafdfasdf2

CLI also returns the error on oxide instance nic list

stevenkaram@MacBook-Pro-2 oxide.rs % oxide instance nic list --instance fdsafdfasdf2 --profile oxide2 --project skaram
[]
Error Response: status: 400 Bad Request; headers: {"content-type": "application/json", "x-request-id": "d0ac6b62-81ca-49fc-b6e5-a4ca4faa4768", "content-length": "414", "date": "Wed, 21 Jan 2026 02:34:30 GMT"}; value: Error { error_code: Some("InvalidRequest"), message: "The network interface with ID '721f9da1-1d04-4809-89c0-c87520b39af1' is a dual-stack NIC, with IPv4 address '172.30.0.6' and IPv6 address 'fd2a:c6e9:d592::6'. However, the version of the client being used is unable to fully represent both IPv4 and IPv6 addresses. Update your client and retry the request.", request_id: "d0ac6b62-81ca-49fc-b6e5-a4ca4faa4768" }

[sion42x]

@ahl fairly certain it's a dup as you mentioned. Since this issue results in CLI error as well should I add the latest comment to the original issue?

[sion42x]

On second thought it might be a separate issue; failed addition of a network adapter to an instance results in the instance details being unviewable in console and NIC view failing in CLI

[bnaecker]

This is all down to the same underlying issue when adding dual-stack IP support in Omicron. The failure to create the instance is because the new default parameter for creating an instance's network interface is "dual-stack", which unhelpfully requires an IPv6 IP pool. If you've not set one up, the default isn't usable. Bad UX, sorry about that.

The failure to receive the data indicates that you've somehow managed to successfully create an instance, which now has a dual-stack network interface, but that the older client you're using can't faithfully represent both the IPv4 and IPv6 addresses. I honestly wasn't sure what to do here, but probably chose the wrong thing. We could have returned the IPv4 address and ignored the IPv6 one. I opted not to do that because it would silently lose information. Your instance has an IPv6 address, but you'd never know it. You can't communicate "partial success" with HTTP.

I think the most useful fix here is to add a better default for the network interface during instance creation, which only makes an automatically-chosen IP address for the IP Pools you actually have.