Deleting default-backend doesn't update backend configuration

[Sea-you]

Description of the problem

It looks like that even with the latest 0.11 beta image, whenever you delete the default-backend pod and it comes up with a new IP address, this is not reflected in the configuration, all other updates are working just fine. Additionally, it'd be nice to be able to configure default-backend's backend protocol independently (or force it to h1) as now - after enabling http/2 backend - I have problems with it as it's a simple HTTP/1.1 webserver.

Before, all good:

haproxy-ingress-0-default-backend-5956896444-99ntl                1/1     Running   0          82m   10.232.6.16

backend _default_backend
    mode http
    balance roundrobin
    http-request set-header X-Original-Forwarded-For %[hdr(x-forwarded-for)] if { hdr(x-forwarded-for) -m found }
    http-request del-header x-forwarded-for
    option forwardfor
    server srv001 10.232.6.16:8080 weight 1 proto h2
    server srv002 127.0.0.1:1023 disabled weight 1 proto h2
    server srv003 127.0.0.1:1023 disabled weight 1 proto h2
    server srv004 127.0.0.1:1023 disabled weight 1 proto h2

After deleting pod:

haproxy-ingress-0-default-backend-5956896444-jpnpv                1/1     Running   0          2m26s   10.232.8.19

backend _default_backend
    mode http
    balance roundrobin
    http-request set-header X-Original-Forwarded-For %[hdr(x-forwarded-for)] if { hdr(x-forwarded-for) -m found }
    http-request del-header x-forwarded-for
    option forwardfor
    server srv001 10.232.6.16:8080 weight 1 proto h2
    server srv002 127.0.0.1:1023 disabled weight 1 proto h2
    server srv003 127.0.0.1:1023 disabled weight 1 proto h2

Expected behavior

Default-backend's IP address should be updated in the configuration.

Steps to reproduce the problem

1. Start controller
2. Delete default-backend pod

Environment information

HAProxy Ingress version: v0.11-beta2

Command-line options:

--default-backend-service=ingress/haproxy-ingress-0-default-backend 
--default-ssl-certificate=ingress/tls
--rate-limit-update=0.1 
--verify-hostname=false

Global options:

    backend-protocol: "h2"
    backend-server-slots-increment: "42"
    health-check-interval: "" # Disable backend healthcheck
    max-connections: "20000"
    nbthread: "4"
    ssl-engine: "rdrand"
    ssl-options: "no-tlsv10 no-tlsv11 no-tls-tickets"
    ssl-redirect-code: "301"
    timeout-client-fin: "5s"
    timeout-server-fin: "5s"
    use-htx: "true"

Thanks for all the help in advance!

[jcmoraisjr]

Hi, thank you for the report. In fact v0.11 introduced this issue due to its nature (expose without and ingress) and how v0.11 updates the internal state (tracking everything under ingress objects). A new v0.11 beta should be tagged in a few more days.

[jcmoraisjr]

Just to let you know that v0.11-beta.3 was just tagged, hope that it fixes the issue in your env. Let me know if anything goes wrong.