[multicast] Complete multicast background reconciliation and designated-forwarder election - #11128
Open
zeeshanlakhani wants to merge 11 commits into
Open
[multicast] Complete multicast background reconciliation and designated-forwarder election#11128zeeshanlakhani wants to merge 11 commits into
zeeshanlakhani wants to merge 11 commits into
Conversation
…ation via DDM Designated-forwarder election places a group's external NAT entry on a single switch chosen by group-UUID hash, so external multicast ingresses exactly once. The sled-agent `DdmReconciler` gains a multicast-origin arm that originates underlay member prefixes, with advertise/withdraw methods in ddm-admin-client. `PortManager` exposes the subscription set and a change notification for that reconciler, and Nexus dataplane and member background tasks pick up the elected-switch plumbing.
…esh fixes Sled-agent: - The `instance_manager` awaits every instance's external-IP refresh and folds per-instance errors instead of stopping directly at the first one, so one failed refresh does not leave remaining ports on stale IGW keying. - `refresh_external_ips_inner` treats an instance without a primary NIC as a successful no-op. Such an instance has no OPTE port to refresh, and propagating a `NoPrimaryNic` error would permanently fail `set_eip_gateways` for the whole sled. Note: A regression test registers a NIC-less instance and asserts the sled-wide refresh succeeds. Nexus: - `member_attach`'s source-union cap check skips members already in a non-"Left" state, so a repeat reattach cannot trip the cap. - The VPC IGW mapping query for sled firewall resolution returns probe NIC mappings for probes on that sled. - Reconciler drift telemetry: `entries_misplaced` and `groups_reelected` counters flow through the background status and omdb. - `update_groups` keys DPD updates on the immutable DB tag rather than a rename parameter.
…erage
- `ip_pools`: rewrite multicast range validation as a table-driven test
with an `Expect::{Ok, Reject}` case list, adding IPv6 coverage
(ff02::/16 rejected, ff05::/16 accepted) alongside the v4
reserved-range cases.
- omdb: fold the pools test into test_omdb_multicast_commands, which
exercises pools, groups, members, and info against one setup.
- multicast integration tests cleanup and consolidation.
…::/12 `is_ssm_address` checks the sixteen per-scope ff3x::/32 blocks (`IPV6_SSM_SUBNETS`), but several comments and doc strings described the IPv6 range as the broader ff30::/12 prefix. Align them: the `is_ssm_address` doc cites RFC 4607 for IPv4 232.0.0.0/8 and points at `IPV6_SSM_SUBNETS` for the v6 blocks, and the datastore and join-endpoint docs follow the same framing. Doc-string changes in nexus-external-api flow into the generated OpenAPI documents, so the three non-blessed nexus specs and the latest pointer are regenerated.
…te fixes This includes: - `MulticastDataplaneClient::new` takes only (resolver, log) again, reverting the DB switch-table completeness check. The switch table has no writer, so the DNS-derived client map is authoritative. An empty map makes the pass skip and retry rather than fail against a stale denominator. - `remove_groups` fails when the client map is empty or Dendrite discovery was partial instead of succeeding vacuously, so cleanup cannot delete DB rows while switch entries may remain intact. - Warn when fewer DPD switches report slots than DNS advertised, tolerating a partial slot map. - The external drift check returns early when the incumbent fetch errors rather than falling through to the hash election. Electing without an observed owner could move ownership off an unobserved switch and leave both switches replicating the group's traffic. - The drift check also compares DPD forwarding state (NAT target, VLAN) against what Nexus programs, alongside tag and sources. - Probe create rejects IPv6 multicast memberships in every request form and defaults the by-name auto-create pool hint to V4, since the in-zone joiner cannot pin an IPv6 interface scope yet. A rejected spec or failed probe creation rolls back any group implicitly created during resolution. - Assorted doc and test cleanup.
zeeshanlakhani
changed the title
zeeshanlakhani
changed the title
zeeshanlakhani
force-pushed
the
zl/mcast-build
branch
from
August 25, 2026 05:34
42eb96e to
64bbc8e
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Nexus now elects and reconciles a single external forwarder per multicast group while continuing to rely on the existing DDM/mg-lower ownership of underlay membership origination and propagation.
This includes: