lag: a bundle carries eight active links, and a LACP group's surplus is hot-standby #2024

Merged
jaylouisw merged 1 commit from rev/1118-1942 into main 2026-11-18 20:07:44 +00:00
Owner

One question: how many links can one EtherChannel carry, and what happens to the ports that do not fit?

Verdict, from the vendor text (Cisco, Configuring EtherChannels, Catalyst 2960-X 15.2(3)E and the 15.2(2)E / IE 4010 / 4500 / 3850 chapters):

  • "Each EtherChannel can consist of up to eight compatibly configured Ethernet ports."
  • "For a PAgP EtherChannel, you can configure up to eight ports ... For a LACP EtherChannel, you can configure up to 16 Ethernet ports ... Up to eight ports can be active, and up to eight ports can be in standby mode."
  • "Only eight LACP links can be active at one time ... a link that is in the hot-standby mode becomes active in its place", and the flag is H in show etherchannel summary.
  • The active set is the lowest-numbered members ("hot-standby ports that have lower port numbers become active in the channel first").

PacketForge capped nothing: ten channel-group lines made a ten-member bundle, all ten read (P), the hash was taken across all ten, and the H - Hot-standby (LACP only) line of the summary's own key was unreachable.

Change

  • engine/lag.js: MAX_ACTIVE_LINKS = 8, MAX_MEMBERS_LACP = 16, MAX_MEMBERS_STATIC = 8, maxMembers(), and carryingMembers / activeMembers / standbyMembers; eligibleMembers() (and so selectMember()) now only ever sees the active eight, chosen in port order (comparePortNames, not string order).
  • engine/cli.js: a ninth mode on member and a seventeenth LACP member are refused by name with the number they applied; a member that lands beyond the eighth gets a hot-standby receipt; show etherchannel summary prints (H) for parked members, (s) for a static surplus that only a saved topology can produce, keeps (P) for the eight, and no longer calls a full-width bundle "running on N of M".
  • engine/simulator.js: the lag event's degraded flag measures capacity, not cable count.
  • test/etherchannel-member-limit.test.js (30 assertions): the pure rule, both refusals, the receipt, the letters, the data plane never hashing onto a parked link, and a standby being promoted when an active cable is pulled.
  • research/etherchannel-member-limit.md: sources, what is inferred, what is unresolved (the box's own wording for the ninth/seventeenth line, lacp max-bundle, per-vendor maxima).
  • Docs: docs/CLI.md section on the cap and (H), docs/SIMULATION.md LAG rows, CHANGELOG entry, and docs/BGP-POLICY.md _help() anchors moved by this diff.

Gates

  • node scripts/syntax-check.js: 422 files, all OK.
  • node scripts/check-doc-citations.js: no judged citation more than 40 lines from its code.
  • npm test: 348 suites, 12,065 assertions passed, 0 failed (194 s), including the new suite.
One question: **how many links can one EtherChannel carry, and what happens to the ports that do not fit?** Verdict, from the vendor text (Cisco, *Configuring EtherChannels*, Catalyst 2960-X 15.2(3)E and the 15.2(2)E / IE 4010 / 4500 / 3850 chapters): - "Each EtherChannel can consist of up to eight compatibly configured Ethernet ports." - "For a PAgP EtherChannel, you can configure up to eight ports ... For a LACP EtherChannel, you can configure up to 16 Ethernet ports ... Up to eight ports can be active, and up to eight ports can be in standby mode." - "Only eight LACP links can be active at one time ... a link that is in the hot-standby mode becomes active in its place", and the flag is `H` in `show etherchannel summary`. - The active set is the lowest-numbered members ("hot-standby ports that have lower port numbers become active in the channel first"). PacketForge capped nothing: ten `channel-group` lines made a ten-member bundle, all ten read `(P)`, the hash was taken across all ten, and the `H - Hot-standby (LACP only)` line of the summary's own key was unreachable. ### Change - `engine/lag.js`: `MAX_ACTIVE_LINKS = 8`, `MAX_MEMBERS_LACP = 16`, `MAX_MEMBERS_STATIC = 8`, `maxMembers()`, and `carryingMembers` / `activeMembers` / `standbyMembers`; `eligibleMembers()` (and so `selectMember()`) now only ever sees the active eight, chosen in port order (`comparePortNames`, not string order). - `engine/cli.js`: a ninth `mode on` member and a seventeenth LACP member are refused by name with the number they applied; a member that lands beyond the eighth gets a hot-standby receipt; `show etherchannel summary` prints `(H)` for parked members, `(s)` for a static surplus that only a saved topology can produce, keeps `(P)` for the eight, and no longer calls a full-width bundle "running on N of M". - `engine/simulator.js`: the `lag` event's `degraded` flag measures capacity, not cable count. - `test/etherchannel-member-limit.test.js` (30 assertions): the pure rule, both refusals, the receipt, the letters, the data plane never hashing onto a parked link, and a standby being promoted when an active cable is pulled. - `research/etherchannel-member-limit.md`: sources, what is inferred, what is unresolved (the box's own wording for the ninth/seventeenth line, `lacp max-bundle`, per-vendor maxima). - Docs: `docs/CLI.md` section on the cap and `(H)`, `docs/SIMULATION.md` LAG rows, CHANGELOG entry, and `docs/BGP-POLICY.md` `_help()` anchors moved by this diff. ### Gates - `node scripts/syntax-check.js`: 422 files, all OK. - `node scripts/check-doc-citations.js`: no judged citation more than 40 lines from its code. - `npm test`: **348 suites, 12,065 assertions passed, 0 failed** (194 s), including the new suite.
lag: a bundle carries eight active links, and a LACP group's surplus is hot-standby (#2023)
All checks were successful
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 4s
CI / Classify the change (pull_request) Successful in 6s
CI / Server smoke test (pull_request) Successful in 10s
CI / Docker build (pull_request) Successful in 23s
CI / Test (Node 20) (pull_request) Successful in 52s
CI / Test (Node 22) (pull_request) Successful in 55s
CI / Test (Node 18) (pull_request) Successful in 58s
CI / Arm auto-merge (pull_request) Successful in 32s
4ea80e3f4f
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-18 20:00:24 +00:00
jaylouisw deleted branch rev/1118-1942 2026-11-18 20:07:45 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!2024
No description provided.