upstream hash by host

[LoicMahieu]

I am not completely confident that it is a bug report, do not hesitate to contradict me ;)

I have several ingresses and for some of them, I need to stick to one backend.
The reason is that theses apps are pretty old JSP apps that consumes a lot of memory. By arranging to gather requests from the same HTTP host on the same backend, it saves me some memory.

At the time of the 0.20.0, I put this annotation:

"nginx.ingress.kubernetes.io/upstream-hash-by": "$host"

It seems to worked well. But I recently upgraded to 0.22.0 from 0.20.0 and now the requests are balanced to each backends.

• What's your opinion about upstream-hash-by = "$host" ? Is there a good way ? Will the controller switch to another backend if the previous is not healthy ?

In the 0.20.0, I observed some errors (113: No route to host) while connecting to upstream due to backend scale up/down.

• Is there a need to specify upstream-hash-by-subset and upstream-hash-by-subset-size ?
  It tried it but not works.

Thanks for your help. 👍

NGINX Ingress controller version: 0.22.0
Kubernetes version: v1.11.6-gke.3
Cloud provider or hardware configuration: GKE
Install tools: Helm

[JordanP]

You seem to be describing several problems.

Could you tell me more about In the 0.20.0, I observed some errors (113: No route to host) while connecting to upstream due to backend scale up/down. ? I might be seeing something like this too.

[LoicMahieu]

Hi @JordanP
Thanks for your response. Sorry, indeed my message is not clear.

The primary problem that I want to solves is: stick requests from the same http-host to the same backend.
For this reason, I added this annotation to the desired ingress:

"nginx.ingress.kubernetes.io/upstream-hash-by": "$host"

In 0.20.0, it worked well, in my tests. But recently I tried to scale up the backends and then scale down. After that the ingress-controller was failing to proxy the requests. Logs was: (113: No route to host) while connecting to upstream. A restart of controller pods solved the problem.

In 0.22.0, it seems that requests are not sticked to the same backend, depending on the http-host. I tried scale up/down and not more routing errors.

[diegows]

Hi @LoicMahieu, could you share all the annotations (or the whole ingress object) you are using in the ingress? I've modified some code that's related to upstream-hash-by annotation, maybe I broke some specific use case.

If you don't need the "subset" feature, you don't need to add those annotations. It's disabled by default.

[LoicMahieu]

Hi @diegows

Annotations are quite simple:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  annotations:
    kubernetes.io/ingress.class: nginx
    nginx.ingress.kubernetes.io/from-to-www-redirect: "true"
    nginx.ingress.kubernetes.io/upstream-hash-by: $host
  name: app-somedomain
spec:
  rules:
  - host: somedomain.com
    http:
      paths:
      - backend:
          serviceName: backend
          servicePort: 80
        path: /

[diegows]

You must use quotes in all the annotations or K8s API is going to drop it somewhere before reaching the Nginx controller.

[LoicMahieu]

Value is quoted in our sources. I used kubectl get -o yaml and it prints it without quotes ;)

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[fejta-bot]

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/close

[k8s-ci-robot]

@fejta-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.