Skip to content

smartcontract,controlplane: add flex-algo link/tenant CLI extensions and migrate command#3513

Merged
elitegreg merged 11 commits intomainfrom
bc/rfc18-pr3
Apr 22, 2026
Merged

smartcontract,controlplane: add flex-algo link/tenant CLI extensions and migrate command#3513
elitegreg merged 11 commits intomainfrom
bc/rfc18-pr3

Conversation

@ben-malbeclabs
Copy link
Copy Markdown
Contributor

@ben-malbeclabs ben-malbeclabs commented Apr 10, 2026

RFC-18 flex-algo · PR 3 of 5 · see rfcs/rfc-0018-flex-algo.md
Depends on: #3512 (PR 2)
Series: #3497 · #3512 · #3513 · #3514 · #3515

Summary of Changes

  • Extends doublezero link get/list/update to display topology assignments (link_topologies) and drain status; adds --link-topology (comma-separated topology names, default to clear all) and --unicast-drained flags to link update; adds --topology filter to link list
  • Extends doublezero tenant get/list/update to display included topologies (include_topologies) and adds --include-topologies (comma-separated topology names, default to clear) to tenant update
  • Adds doublezero-admin migrate flex-algo [--dry-run] command that backfills link topology assignments and VPNv4 loopback flex-algo node segments across all existing devices and links

Diff Breakdown

Category Files Lines (+/-) Net
Core logic 11 +471 / -57 +414
Scaffolding 7 +16 / -5 +11

Mostly core logic — the migrate command alone is 180 lines of backfill and dry-run orchestration.

Key files (click to expand)
  • controlplane/doublezero-admin/src/cli/migrate.rs — new: migrate flex-algo command; validates UNICAST-DEFAULT PDA exists, backfills link topologies for all links, backfills flex-algo node segments for all VPNv4 loopback interfaces on all devices; supports --dry-run
  • smartcontract/cli/src/link/list.rs — adds --topology <name> filter and displays link_topologies/unicast_drained columns, resolving topology pubkeys to human-readable names
  • smartcontract/cli/src/link/get.rs — displays topology assignments and drain status; fetches topology map to resolve pubkeys to names
  • smartcontract/cli/src/link/update.rs — adds --link-topology (accepts comma-separated list of topology names; use default to clear) and --unicast-drained flags
  • smartcontract/cli/src/tenant/list.rs — displays include_topologies column
  • smartcontract/cli/src/tenant/get.rs — displays included topology names
  • smartcontract/cli/src/tenant/update.rs — adds --include-topologies (accepts comma-separated list; use default to clear)

Testing Verification

  • cargo test -p doublezero_sdk -p doublezero_cli -p doublezero-admin passes
  • cargo clippy -- -D warnings clean
  • make rust-fmt applied before commit

@ben-malbeclabs ben-malbeclabs force-pushed the bc/rfc18-pr3 branch 2 times, most recently from 6097a83 to 254484a Compare April 15, 2026 00:43
@ben-malbeclabs ben-malbeclabs marked this pull request as ready for review April 15, 2026 04:00
elitegreg added a commit that referenced this pull request Apr 18, 2026
…y processors (#3497)

RFC-18 flex-algo · PR 1 of 5 · see `rfcs/rfc-0018-flex-algo.md`
Series: #3497 · #3512 · #3513 · #3514 · #3515

## Summary of Changes
- Adds `TopologyInfo` onchain account (RFC-18 flex-algo) with
`admin_group_bit`, `flex_algo_number`, and `TopologyConstraint`; creates
`AdminGroupBits` resource extension singleton for bit allocation
- Adds four new foundation-only instructions (107–110):
`CreateTopology`, `DeleteTopology`, `ClearTopology`, `BackfillTopology`
- Extends `Link` with `link_topologies` (Vec<Pubkey>) and `link_flags`
(LINK_FLAG_UNICAST_DRAINED); extends `Tenant` with `include_topologies`;
adds `flex_algo_node_segments` to Interface V2
- `CreateLink` now requires the unicast-default topology account and
auto-tags the link into it
- `UpdateLink` gains foundation-gated `link_topologies` update and
contributor-gated `unicast_drained` flag

## Diff Breakdown
| Category    | Files | Lines (+/-)   | Net   |
|-------------|-------|---------------|-------|
| Core logic  |    22 | +1,306 / -38  | +1,268|
| Scaffolding |    49 | +133 / -3     |  +130 |
| Tests       |    27 | +4,592 / -74  | +4,518|

The scaffolding is mechanical struct field additions (`link_topologies:
vec![]`, `include_topologies: vec![]`, `flex_algo_node_segments:
vec![]`) across CLI, SDK, activator, and client to keep the workspace
compiling. Core logic is concentrated in the new topology processors,
state types, and link/activate updates.

<details>
<summary>Key files (click to expand)</summary>

-
`smartcontract/programs/doublezero-serviceability/src/state/interface.rs`
— updates Interface V2 (discriminant 1) to include
`flex_algo_node_segments: Vec<FlexAlgoNodeSegment>`; V1 accounts
(pre-CYOA format) are left as-is; pre-RFC-18 V2 accounts on mainnet are
upgraded in-place by `MigrateDeviceInterfaces` before this
deserialization path is used
-
`smartcontract/programs/doublezero-serviceability/src/state/topology.rs`
— new `TopologyInfo` account: `admin_group_bit` (u8), `flex_algo_number`
(128+bit), `TopologyConstraint` (IncludeAny/Exclude)
-
`smartcontract/programs/doublezero-serviceability/src/processors/topology/`
— five new processors: create (allocates AdminGroupBit, validates
IdRange 1–127), delete, clear (removes topology from all links),
backfill (allocates FlexAlgoNodeSegment on all device interfaces)
-
`smartcontract/programs/doublezero-serviceability/src/processors/link/update.rs`
— adds foundation-gated `link_topologies` update (validates each
topology account onchain) and contributor-gated `unicast_drained` flag
(LINK_FLAG_UNICAST_DRAINED bit 0)
-
`smartcontract/programs/doublezero-serviceability/src/processors/link/activate.rs`
— requires `unicast_default_topology_account` as a mandatory account;
auto-tags new link into the unicast-default topology on activation
-
`smartcontract/programs/doublezero-serviceability/src/processors/globalconfig/set.rs`
— creates `AdminGroupBits` ResourceExtension PDA on global config
initialization
- `smartcontract/programs/doublezero-serviceability/src/state/link.rs` —
adds `link_topologies: Vec<Pubkey>`, `link_flags: u64`, and
`LINK_FLAG_UNICAST_DRAINED = 0x01`
- `smartcontract/programs/doublezero-serviceability/src/instructions.rs`
— registers four new instruction variants (CreateTopology=107,
DeleteTopology=108, ClearTopology=109, BackfillTopology=110)

</details>

## Testing Verification
- `make rust-lint` and `make rust-test` pass clean on this branch
- `topology_test.rs` (~2,400 lines) covers all four topology processors:
create/delete/clear/backfill, including error paths (duplicate names,
out-of-range bits, unauthorized signers, invalid topology accounts),
segment allocation, and multi-topology backfill
- Existing `link_wan_test.rs`, `tenant_test.rs`, and telemetry tests
updated to account for the new mandatory
`unicast_default_topology_account` in `ActivateLink`

---------

Co-authored-by: Greg Mitchell <greg@malbeclabs.com>
@elitegreg elitegreg force-pushed the bc/rfc18-pr3 branch 2 times, most recently from e3e70f5 to b15b100 Compare April 21, 2026 20:15
ben-malbeclabs added a commit that referenced this pull request Apr 21, 2026
RFC-18 flex-algo · PR 2 of 5 · see `rfcs/rfc-0018-flex-algo.md`
Series: #3497 · #3512 · #3513 · #3514 · #3515

## Summary of Changes
- Adds five `doublezero link topology` subcommands: `create`, `delete`, `clear`, `backfill`, `list`
- `topology create` allocates the admin-group bit and flex-algo number onchain, then automatically backfills `FlexAlgoNodeSegment` entries on all activated Vpnv4 loopback interfaces
- `topology backfill` and `topology clear` batch at 16 accounts per transaction to stay within Solana's 32-account limit; zero accounts → zero transactions sent
- `topology list` displays each topology's name, admin-group bit, flex-algo number, BGP color community, constraint, and count of tagged links
- Backs each subcommand with a Rust SDK command in `doublezero_sdk::commands::topology`
- Simplifies the `BackfillTopology` onchain processor: removes the two-pass SR ID pre-marking loop; keeping the `SegmentRoutingIds` resource in sync with off-chain allocations is the activator's responsibility

## Diff Breakdown
| Category    | Files | Lines (+/-)    | Net    |
|-------------|-------|----------------|--------|
| Core logic  |    11 | +1,572 / -32   | +1,540 |
| Scaffolding |     8 |   +91 /  -6    |   +85  |
| Tests       |     1 |   +25 / -16    |    +9  |

Core logic includes inline unit tests in the SDK command files.

<details>
<summary>Key files (click to expand)</summary>

- `smartcontract/sdk/rs/src/commands/topology/create.rs` — creates topology, then enumerates devices and calls `BackfillTopologyCommand` for those with Vpnv4 loopbacks; returns `CreateTopologyResult { signature, topology_pda, backfill_signatures }`
- `smartcontract/sdk/rs/src/commands/topology/backfill.rs` — batches devices in chunks of 16; returns `Vec<Signature>`; empty device list sends no transactions
- `smartcontract/sdk/rs/src/commands/topology/clear.rs` — batches links in chunks of 16; returns `Vec<Signature>`; empty link list sends no transactions
- `smartcontract/sdk/rs/src/commands/topology/list.rs` — fetches all `TopologyInfo` accounts and cross-references link accounts to compute per-topology link counts
- `smartcontract/cli/src/topology/create.rs` — CLI wrapper; prints PDA and backfill transaction count
- `smartcontract/cli/src/topology/list.rs` — tabular display of name, bit, algo number, BGP color, constraint, link count
- `smartcontract/programs/doublezero-serviceability/src/processors/topology/backfill.rs` — removes two-pass SR ID pre-marking; single-pass allocation only

</details>

## Testing Verification
- `cargo test -p doublezero_sdk -p doublezero_cli` passes
- `cargo clippy -- -D warnings` clean
- `make rust-fmt` applied
Base automatically changed from bc/rfc18-pr2 to main April 21, 2026 20:40
Copy link
Copy Markdown
Contributor

@elitegreg elitegreg left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

a few minor things to fix

Comment thread smartcontract/cli/src/tenant/get.rs Outdated
Comment thread smartcontract/cli/src/link/list.rs Outdated
Comment thread smartcontract/cli/src/link/list.rs Outdated
Comment thread smartcontract/cli/src/link/get.rs Outdated
ben-malbeclabs and others added 7 commits April 22, 2026 17:27
topology create/delete/clear/backfill and link update all normalize the
user-supplied topology name to lowercase before deriving PDAs or comparing
against stored names. clear and link update use case-insensitive comparison
to handle existing uppercase data during transition.
@elitegreg elitegreg enabled auto-merge (squash) April 22, 2026 17:27
@elitegreg elitegreg merged commit 855a28c into main Apr 22, 2026
36 checks passed
@elitegreg elitegreg deleted the bc/rfc18-pr3 branch April 22, 2026 19:07
elitegreg pushed a commit that referenced this pull request Apr 23, 2026
RFC-18 flex-algo · PR 4 of 5 · see
[`rfcs/rfc-0018-flex-algo.md`](../tree/main/rfcs/rfc-0018-flex-algo.md)
Depends on: #3512 (PR 2), #3513 (PR 3)
Series: #3497 · #3512 · #3513 · #3514 · #3515

## Summary of Changes

- Python SDK: deserialize `TopologyConstraint`, `TopologyInfo`, and
`FlexAlgoNodeSegment`; read `flex_algo_node_segments` from the RFC-18
`InterfaceV3` account format
- TypeScript SDK: deserialize `FlexAlgoNodeSegment` from `InterfaceV3`
account format; guard the segments loop against pre-RFC-18 mainnet
accounts where the segment count reads garbage bytes
- SDK fixtures: regenerate device, link, tenant, and user fixtures to
include the new `InterfaceV3` fields
- TypeScript RPC client: remove the `AbortController` timeout wrapper
around `fetch` (was surfacing as spurious `TimeoutError` on slow
`getProgramData` responses); bump the compat test's request timeout to
120s
- `tryReadString` in `borsh-incremental`: check remaining buffer length
before reading to avoid out-of-bounds reads
- CHANGELOG / e2e compatibility test: document the mandatory CLI upgrade
boundary for RFC-18 `InterfaceV3` (devices written by the new program
cannot be deserialized by pre-RFC-18 CLI versions)

Note: activator changes that were originally planned for this PR have
been dropped — the doublezero activator is deprecated.

## Diff Breakdown

| Category   | Files | Lines (+/-) |
|------------|-------|-------------|
| Python SDK |     1 | +61 / -1    |
| TS SDK     |     3 | +27 / -3    |
| Fixtures   |     5 | +4 / -1 + bin |
| Docs/e2e   |     2 | +2 / -2     |

## Testing Verification

- Python SDK: `uv run pytest` passes under `sdk/serviceability/python/`,
including fixture deserialization
- TypeScript SDK: `bun test` passes under
`sdk/serviceability/typescript/`, including compat fixture
deserialization with the new `InterfaceV3` fields
- `cargo check --workspace` clean
ben-malbeclabs added a commit that referenced this pull request Apr 29, 2026
…troller support

RFC-18 flex-algo · PR 5 of 5 · see `rfcs/rfc-0018-flex-algo.md`
Depends on: #3497 (PR 1) — does not require PRs 2–4; can merge after PR
1
Series: #3497 · #3512 · #3513 · #3514 · #3515

## Summary of Changes
- Adds `TopologyInfo` account type to the Go SDK (`state.go`,
`deserialize.go`, `client.go`), including `TopologyType`,
`TopologyConstraint`, `LinkFlagUnicastDrained`,
`LinkTopologies`/`LinkFlags` on `Link`, and `IncludeTopologies` on
`Tenant`
- Auto-loads `/etc/doublezero-controller/features.yaml` at startup if
present (silently skips if absent); when `flex_algo.enabled: true`,
populates topology data into the state cache, resolves tenant color
communities via `resolveTenantColors`, and emits IS-IS flex-algo node
segment and link configuration into the Arista EOS template
- Template emits correct flex-algo definitions for both `include-any`
constraint (`administrative-group include any N exclude 0`) and
`exclude` constraint (`administrative-group exclude N,0`) topologies;
gated per-link via `link_tagging.exclude` and per-tenant via
`community_stamping`
- Updates `e2e/compatibility_test.go` to set the mandatory upgrade
boundary for RFC-18 interface/link operations to `before: "0.18.0"`

## Diff Breakdown
| Category    | Files | Lines (+/-)  | Net   |
|-------------|-------|--------------|-------|
| Core logic  |     6 | +368 / -6    | +362  |
| Scaffolding |     2 | +22 / -0     |  +22  |
| Tests       |     4 | +482 / -2    | +480  |
| Fixtures    |     2 | +415 / -0    | +415  |
| Docs        |     1 | +4 / -0      |   +4  |

Most of the weight is tests and fixtures; core logic is the controller
cache/template and Go SDK deserialization.

> **Note on diff size:** The full diff vs `main` is large because this
branch is stacked on PR 1 (#3497), which has not yet merged. The table
above reflects only PR 5's contribution. After PR 1 merges and this
branch is rebased, the diff shrinks to the ~15 files shown here.

<details>
<summary>Key files (click to expand)</summary>

-
[`controlplane/controller/internal/controller/server.go`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-f5e715ed1377ac408a4ecde2c60cbcb98327512c7d1496114ef143cb27e7afaf)
— populates `Topologies` map in state cache; resolves `LinkTopologies`
pubkeys to topology names and `UnicastDrained` from `LinkFlags`;
computes `TenantTopologyColors`; `resolveTenantColors` falls back to
`unicast-default` when tenant has no explicit topology assignments
-
[`controlplane/controller/internal/controller/templates/tunnel.tmpl`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-d67a980326aca35fdabdd5445fbbab4716f1ae412af4540770d5a9ee116b04d9)
— IS-IS flex-algo node segment config, link admin-group tagging, BGP
`next-hop resolution ribs` and `set extcommunity` stamping, and full
rollback (`no` commands) when disabled; handles both `include-any` and
`exclude` constraint topologies
-
[`controlplane/controller/internal/controller/features_config.go`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-24f09aab331b4d309662e5ee6550fd38fa173676ae868db381d9af23cf645f1d)
— `FeaturesConfig` struct with `flex_algo.enabled`,
`link_tagging.exclude`, and `community_stamping` controls; auto-loaded
from `/etc/doublezero-controller/features.yaml`
-
[`controlplane/controller/internal/controller/models.go`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-bdc4e34592a7435dcde7dab78ffd735dcb9847447f0113bd9f3afc8d3d01f864)
— adds `LinkTopologies`, `UnicastDrained`, `FlexAlgoNodeSegments`,
`TenantTopologyColors` to controller model types; `TopologyModel` and
`FlexAlgoEnabled()` for template rendering
-
[`smartcontract/sdk/go/serviceability/state.go`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-057f627e5f2629b35f1976fba6eaa1b3c507c511b0f47d2f80a3a1d0c684c4a1)
— adds `TopologyType`, `TopologyConstraint`, `TopologyInfo`,
`LinkFlagUnicastDrained`; extends `Link` with
`LinkTopologies`/`LinkFlags` and `Tenant` with `IncludeTopologies`
-
[`smartcontract/sdk/go/serviceability/deserialize.go`](https://github.com/malbeclabs/doublezero/pull/3515/files#diff-b71f90b1b3e00de9bd820920e63f5615edb9b7d110c2b39f38895847509a7844)
— adds `DeserializeTopologyInfo`; extends `DeserializeLink` with
`LinkTopologies`/`LinkFlags` and `DeserializeTenant` with
`IncludeTopologies`

</details>

## Testing Verification
- `TestGetConfig_FlexAlgo` (new): render tests for flex-algo disabled,
enabled with both constraint types (`include-any` and `exclude`), and
link excluded from tagging — all pass
- `Test_resolveTenantColors` (new): 5 table-driven cases covering
fallback to `unicast-default`, missing topology, single/multiple known
pubkeys, and unknown pubkey — all pass
- `TestE2E_IBRL`, `TestE2E_IBRL_WithAllocatedAddr`, `TestE2E_Multicast`
— all pass; flex-algo config is suppressed in e2e (no `features.yaml`
present)
- `exclude` constraint flex-algo syntax validated on physical lab
switches (chi-dn-dzd5–dzd8)

---------

Co-authored-by: Greg Mitchell <greg@malbeclabs.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants