tcp-balancer: HAProxy 3.3 breaks due to frontend/backend name collision

Problem

The tcp-balancer app fails to start after HAProxy was updated to version 3.3. The app uses haproxy:latest which now pulls a version that no longer supports having frontend and backend with identical names.

Error Message

[ALERT]    (8) : config : Parsing [/usr/local/etc/haproxy/haproxy.cfg:18]: backend 'http' has the same name as frontend 'http' declared at /usr/local/etc/haproxy/haproxy.cfg:12. This is no longer supported as of 3.3. Please rename one or the other.

Root Cause

The configmap template in packages/apps/tcp-balancer/templates/configmap.yaml creates frontend/backend pairs with matching names:

frontend http / backend http
frontend https / backend https
frontend kubernetes / backend kubernetes
frontend talos / backend talos

This was valid in HAProxy < 3.3 but is now a breaking error.

Suggested Fix

Rename backends to use a suffix (e.g., backend http_backend, backend https_backend, etc.)
Pin the image version to avoid unexpected breakage from :latest (e.g., haproxy:2.9 or haproxy:3.2)

Option 2 is recommended as a best practice regardless—using :latest in production workloads can lead to unexpected breakage like this.

Affected Files

packages/apps/tcp-balancer/templates/configmap.yaml - HAProxy config template
packages/apps/tcp-balancer/templates/deployment.yaml - Image reference (haproxy:latest)

The tcp-balancer app fails with HAProxy 3.3 because the config template defines frontend/backend pairs with identical names (like frontend http / backend http), which is no longer allowed in HAProxy 3.3+ and causes a startup error. The relevant config is in packages/apps/tcp-balancer/templates/configmap.yaml and the image is set to haproxy:latest in deployment.yaml, so the latest (breaking) version is always pulled.

To fix this, rename each backend to use a unique name (for example, backend http_backend instead of backend http) in the configmap template, and update the corresponding default_backend references in the frontend sections. Also, pin the HAProxy image in the deployment template to a known working version (such as haproxy:2.9 or haproxy:3.2) instead of using :latest to avoid future unexpected breakage. This approach is supported by both the code and a relevant GitHub issue discussing explicit version pinning for HAProxy in managed apps like tcp-balancer (source, config example, deployment example).

To reply, just mention @dosu.

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

We hit the same failure on Cozystack 1.6.1: haproxy:latest resolves to 3.4.4 now and the pod dies at config parse on the frontend/backend name collision. #2321 has the right rename but has been idle since July, so #4366 carries it forward with the image pinned to the chart appVersion, the empty-whitelist guard and the unittest suite from that review.

fix(tcp-balancer): rename backends and pin HAProxy so the chart start…

Navigation Menu

Repository navigation

tcp-balancer: HAProxy 3.3 breaks due to frontend/backend name collision #1966

Problem

Error Message

Root Cause

Suggested Fix

Affected Files

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

tcp-balancer: HAProxy 3.3 breaks due to frontend/backend name collision #1966

Description

Problem

Error Message

Root Cause

Suggested Fix

Affected Files

Activity

dosubot commented on Feb 3, 2026

yankawai commented on Sep 22, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions