A LAG keeps hashing traffic onto a member whose link is down #1263

Closed
opened 2026-11-09 20:01:39 +00:00 by jaylouisw · 0 comments
Owner

What happens

Two switches joined by a two-member port-channel. With both cables in, a ping crosses. Take one cable
out
and every ping fails:

both links up      true,true,true,true
one member down    false,false,false,false,false,false
lag picks after:   ["Fa0/3","Fa0/3","Fa0/3","Fa0/3","Fa0/3","Fa0/3"]      <- the dead member, every time
drops:             {"total":7,"byReason":{"link down":7}}

lagLib.selectMember() filtered on enabled and nothing else, and the forwarding path's candidate list
filtered spanning tree and VLAN membership and nothing else. So the hash kept landing on the unplugged member
— and Fa0/2, which was up and idle, carried nothing.

show etherchannel summary reported both members (P) throughout:

Group  Protocol   Ports
1      active     Fa0/2(P) Fa0/3(P)

(P) meaning "this port is carrying traffic", printed for a port whose link was down.

Why it matters

A bundle is supposed to survive losing a member — that is the entire reason to build one. Here, losing a
member cost roughly half the flows (whatever hashed to it) while the surviving member sat idle, and the only
clue was a drop count on a port that looked bundled.

Expected

  • a member that cannot carry a frame is not a candidate: its cable is out, its link is down, its far port is
    shut, it is administratively down, or spanning tree blocks it;
  • the hash is taken over the members that can, so the bundle keeps working while one member is down — and
    with nothing wrong, the selection is byte-for-byte what it was, so flows do not re-balance on an unrelated
    change;
  • when no member can carry, the bundle is down and the frames are still attempted and dropped with a
    reason, so the loss stays countable (#1258) rather than disappearing into a bundle that looks healthy;
  • show etherchannel summary marks (P) only for members that can carry, (D) otherwise, says why in the
    same sentence show interfaces uses, and reports how many members the bundle is running on;
  • the lag timeline event records the eligible members, so a viewer can see why one was not chosen.

Scope

The existing LAG selection rule, the existing forwarding candidate list, and the existing show etherchannel
output. No new commands.

## What happens Two switches joined by a two-member port-channel. With both cables in, a ping crosses. Take **one cable out** and every ping fails: ``` both links up true,true,true,true one member down false,false,false,false,false,false lag picks after: ["Fa0/3","Fa0/3","Fa0/3","Fa0/3","Fa0/3","Fa0/3"] <- the dead member, every time drops: {"total":7,"byReason":{"link down":7}} ``` `lagLib.selectMember()` filtered on `enabled` and nothing else, and the forwarding path's candidate list filtered spanning tree and VLAN membership and nothing else. So the hash kept landing on the unplugged member — and `Fa0/2`, which was up and idle, carried nothing. `show etherchannel summary` reported both members `(P)` throughout: ``` Group Protocol Ports 1 active Fa0/2(P) Fa0/3(P) ``` `(P)` meaning "this port is carrying traffic", printed for a port whose link was down. ## Why it matters A bundle is supposed to survive losing a member — that is the entire reason to build one. Here, losing a member cost roughly half the flows (whatever hashed to it) while the surviving member sat idle, and the only clue was a drop count on a port that looked bundled. ## Expected * a member that cannot carry a frame is not a candidate: its cable is out, its link is down, its far port is shut, it is administratively down, or spanning tree blocks it; * the hash is taken over the members that can, so the bundle keeps working while one member is down — and with nothing wrong, the selection is byte-for-byte what it was, so flows do not re-balance on an unrelated change; * when **no** member can carry, the bundle is down and the frames are still attempted and dropped with a reason, so the loss stays countable (#1258) rather than disappearing into a bundle that looks healthy; * `show etherchannel summary` marks `(P)` only for members that can carry, `(D)` otherwise, says why in the same sentence `show interfaces` uses, and reports how many members the bundle is running on; * the `lag` timeline event records the eligible members, so a viewer can see why one was not chosen. ## Scope The existing LAG selection rule, the existing forwarding candidate list, and the existing `show etherchannel` output. No new commands.
Sign in to join this conversation.
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#1263
No description provided.