Transportserver resource breaks with a long FQDN host value.

[jppitout]

NGINX Ingress controller version: 1.11.3
Kubernetes version: v1.19

• Cloud provider: Azure
• OS: Ubuntu 16.04.7 LTS (Xenial Xerus)
• Kernel: 4.15.0-143-generic
• Install tools: Kubernetes manifests

Description
When adding a transportserver resource with a long FQDN (in our case 57 characters) we get the below error repeated in the logs and nginx-ingress controller takes a very long time to startup affecting other ingress resources:

stderr: "2021/06/24 14:33:38 [emerg] 21#21: could not build map_hash, you should increase map_hash_bucket_size: 64\nnginx: [emerg] could not build map_hash, you should increase map_hash_bucket_size: 64\n"

To Reproduce
Steps to reproduce the behavior:

Install the ingress controller

$ git clone https://github.com/nginxinc/kubernetes-ingress/
$ cd kubernetes-ingress/deployments/common/crds
$ kubectl apply \
  -f k8s.nginx.org_policies.yaml \
  -f k8s.nginx.org_transportservers.yaml \
  -f k8s.nginx.org_virtualserverroutes.yaml \ 
  -f k8s.nginx.org_virtualservers.yaml
$ cd ..
$ kubectl apply \
  -f ns-and-sa.yaml \
  -f default-server-secret.yaml \
  -f ingress-class.yaml \
  -f nginx-config.yaml \
$ cd ../rbac/
$ kubectl apply -f rbac.yaml
$ cd ../deployment/

Add the following to args: in nginx-ingress.yaml spec:

          - -report-ingress-status                                              
          - -external-service=nginx-ingress                                     
          - -ingress-class=nginx                                                
          - -enable-tls-passthrough                                             
          - -enable-custom-resources=true                                       

Apply deployment manifest:

$ kubectl apply -f nginx-ingress.yaml

Install an application that will act as a backend (is just an echo app)
kubectl apply -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/master/docs/examples/http-svc.yaml

Create an transportserver resource:

echo "
  apiVersion: k8s.nginx.org/v1alpha1
  kind: TransportServer
  metadata:
    name: http-svc
    namespace: nginx-ingress
  spec:
    ingressClassName: nginx
    listener:
      name: tls-passthrough
      protocol: TLS_PASSTHROUGH
    host: foo.bar.example.with.a.long.fullyqulaifieddomainname.com
    upstreams:
      - name: http-svc
        service: http-svc
        port: 80
    action:
      pass: http-svc

" | kubectl apply -f -

Expected behavior
Nginx-ingress controller to start without errors.

Is there a way to set map_hash_bucket_size in the configmap? I've not been able to find a way.

Is server-name-hash-bucket-size related? I have tried setting the below but these don't change the behaviour:

  #server-name-hash-bucket-size: "256"
  #server-name-hash-max-size: "1024"
  server-names-hash-bucket-size: "256"
  server-names-hash-max-size: "1024"

Note: When the host value in the transportserver resource has less characters the error goes away.

[github-actions]

Hi @jppitout thanks for reporting!

Be sure to check out the docs while you wait for a human to take a look at this 🙂

Cheers!

[jppitout]

Quick update: Using custom templates with this template I managed to get map_hash_bucket_size 128; set in nginx.conf of the ingress controller pod but am still seeing the below error:

stderr: "2021/06/28 11:52:07 [emerg] 32#32: could not build map_hash, you should increase map_hash_bucket_size: 64\nnginx: [emerg] could not build map_hash, you should increase map_hash_bucket_size: 64\n"

Code snippet from /etc/nginx/nginx.conf in the ingress pod:

    server_names_hash_max_size 1024;
    server_names_hash_bucket_size 256;
    map_hash_bucket_size 512;
    map_hash_max_size 2048;
    variables_hash_bucket_size 256;
    variables_hash_max_size 1024;

Note: Above was inside the http { stanza and above any map stanzas.
It seems that adding to the map_hash settings to the stream context inside http stanza resolves the issue. But I'm still testing...

[jppitout]

For anyone else that may experience similar, in the end I managed to solve this with stream-snippets in the configmap:

  stream-snippets: |
    map_hash_bucket_size 512;
    map_hash_max_size 2048;