[kube-prometheus-stack] Some metrics missing, therefore grafana have empty graphics

andrey-gava commented 2 years ago

Describe the bug a clear and concise description of what the bug is.

Fresh install of chart. Part of Grafana dashboards missing values. For example CPU quota, CPU usage etc. Several dashboards a completely empty (Compute Resources Workload, Networking Namespace Workload) I look in prometheus and there none metrics which have name that starts from "nodenamespace" I tried uninstall and install version 30.0.0 but things are the same

Screens from my external grafana, in the bundled grafana everything the same.

What's your helm version?

3.5.1

What's your kubectl version?

v1.19

Which chart?

kube-prometheus-stack

What's the chart version?

30.0.1

Enter the changed values of values.yaml?

defaultRules:
  create: false
alertmanager:
  enabled: false
grafana:
  enabled: true
prometheus:
  service:
    nodePort: 30090
    type: NodePort
  prometheusSpec:
    podMonitorSelectorNilUsesHelmValues: true
    serviceMonitorSelectorNilUsesHelmValues: false
    remoteWrite:
     - url: http://192.168.3.100:8428/api/v1/write
    externalLabels:
      svc: kubernetes
      project: portal
    retention: 14d
kubeScheduler:
  enabled: false
  serviceMonitor:
    enabled: false
kubeEtcd:
  service:
    port: 2381
    targetPort: 2381
kubeControllerManager:
  enabled: false
  serviceMonitor:
    enabled: false

Enter the command that you execute and failing/misfunctioning.

helm upgrade --install -n monitoring kube-prometheus-stack ./kube-prometheus-stack-30.0.1.tgz -f custom-values.yaml --atomic

avakarev commented 2 years ago

Hi @andrey-gava, try to add

kube-state-metrics:
  releaseLabel: true

To your chart values.

andrey-gava commented 2 years ago

Hi @andrey-gava, try to add
kube-state-metrics:
  releaseLabel: true
To your chart values.

Its already defined inside default values and set to true. https://github.com/prometheus-community/helm-charts/blob/main/charts/kube-prometheus-stack/values.yaml#L1322

altitudems commented 2 years ago

Can confirm same issue here, also pointing to 30.0.1.

My values.yaml

prometheus:
  prometheusSpec:
    retention: 3d
    replicas: 1
    podAntiAffinity: "hard"
    storageSpec:
      volumeClaimTemplate:
        spec:
          storageClassName: longhorn
          accessModes: ["ReadWriteOnce"]
          resources:
            requests:
              storage: 10Gi

    resources:
      limits:
        cpu: "1"
        memory: 2Gi
      requests:
        cpu: 100m
        memory: 1Gi
nodeExporter:
  enabled: true
kubeStateMetrics:
  enabled: true
kube-state-metrics:
  releaseLabel: true
  selfMonitor:
    enabled: false

zedtux commented 2 years ago

I'm facing the same issue. I've added the mentioned option:

diff --git a/deploy/releases/releases.yaml b/deploy/releases/releases.yaml
index 8a1d217..e4d490d 100644
--- a/deploy/releases/releases.yaml
+++ b/deploy/releases/releases.yaml
@@ -163,6 +163,7 @@ releases:
         kube-state-metrics:
           podSecurityPolicy:
             enabled: true
+          releaseLabel: true
         prometheus:
           prometheusSpec:
             podMonitorSelector:

But the deployment didn't show that in its output and I'm still missing the metrics.

I'm using the version 30.1.0 of this chart.

GJim commented 2 years ago

Hi friends, I face the similar situation. (Missing default datasource from grafana)

And I found it's not chart bugs, it's grafana datasource container scrape data via ssl, so edit ssl setting can solve the problem. #1532

↓↓ my grafana datasource container logs

andrey-gava commented 2 years ago

In my case I found what was the problem. When I disabled the defaultRules, I thought that they are only for alertmanager, but they are creatting new metrics names from nested query's. Now when I set it to true, everything works.

defaultRules:
  create: true

altitudems commented 2 years ago

I'm not sure if people are reporting about the same version.

Using the Helm Chart, for version 30.1.0, CPU metrics do not appear in Grafana by default. This is on single k3s node, k8s v1.22. Adding kube-state-metrics.releaseLabel: true or defaultRules.create: true to your values.yml, makes no difference. Those are set by default anyway.

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

gjeusel commented 2 years ago

Any news on this ? I'm facing the same issue

altitudems commented 2 years ago

This is still the case:

K8s Version: v1.22.6+k3s1 Chart Version: 34.1.1

piteros83 commented 2 years ago

I also have the same problem, strange that upgrade from 23.1.1 to 34.8.0 went smoothly, but from 30.1.0 to 34.8.0 apparently not and now I'm missing metrics in Grafana and Lens. What's more, removing and installing Chart again occurs with the same problem regardless the version (removing together with crd's, as it's stated in installation manual). It looks like something still remains on k8s after removing it and not being upgraded by Helm installation.

K8s Version: v1.21.5-eks-bc4871b

EDIT: I found that after uninstalling Helm Chart, it's leaving service object related to kubelet: service/prometheus-stack-kube-prom-kubelet in kube-system namespace

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

zedtux commented 2 years ago

Hilfe

vinfang commented 2 years ago

I'm using chart version kube-prometheus-stack-35.2.0. The prometheus queries show node data like CPU and memory, but none of the container metrics like container_memory_usage_bytes are showing up. Anyone know if it's a setting that would prevent these metrics from showing up in the helm values.yaml or if this would be a bug?

Update: I went back to the prometheus-community/prometheus helm chart 15.10.1, installed that one and I can see all the container metrics on prometheus/thanos to grafana, so I'm not sure what setting in kube-prometheus-stack in the values would make the container metrics not show up.

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

zedtux commented 2 years ago

This issue is still relevant and yes, we agreed, it needs an input from the Prometheus people. :wink:

ff-fgomez commented 2 years ago

I have the same issue, none of the proposed workarounds fix it.

namirsab commented 2 years ago

Same issue here, networking charts dissappeared for me.

namirsab commented 2 years ago

I created a ticket for networking charts #2362

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

bit-project-de commented 2 years ago

I'm having a similar issue but for me it seems to be related to specific nodes. We habe several clusters in multiple environments. Clusters are setup via kubespray. Kube-Prometheus-Stack worked fine some time ago. Now CPU metrics are missing for Pods on specific nodes only. I removed the whole stack, crds, service in kube-system namespace, reinstalled different version, removed all customized values. Still the same nodes won't have metrics for all pods running on them. For example 1 cluster with 3 workers has only worker 2 showing metrics for the pods. Worker 1 and 3 never have metrics.

Any suggestions?

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

vsadanala commented 1 year ago

I have a similar issue lost my metrics post upgrade. Grafana dashboards do not show CPU and memory values .

What Grafana version and what operating system are you using? 9.2

What are you trying to achieve? upgrading kube-prometheus-stack from 10.0.2 to latest version

How are you trying to achieve it? Upgrading step by step from kube-prometheus-stack 10.0.2 to 41.9.1 through helm chart using the official documentation provided by helm => kube-prometheus-stack 41.9.1 · prometheus/prometheus-community

What happened? During the upgrade process from the version 21.x to 22.x, there is an instruction to delete the kube-state-metrics [ kubectl delete deployments.apps -l app.kubernetes.io/instance=prometheus-operator,app.kubernetes.io/name=kube-state-metrics --cascade=orphan]

What did you expect to happen? I lost all my metrics data,even though PVC is mapped to prometheus post upgrade to 22.x, I am not able to see the history data. I am expecting all my historic data should retain post upgrade to latest version

Can you copy/paste the configuration(s) that you are having problems with? all the older data is lost

Did you receive any errors in the Grafana UI or in related logs? If so, please tell us exactly what they were. Grafana dashboards view is also changed post upgrade

Did you follow any online instructions? If so, what is the URL? I followed the below URL for the upgrade process kube-prometheus-stack 41.9.1 · prometheus/prometheus-community

Now i am planning to perform the similar upgrade in my Production environment, but the historical data is very critical and cannot lose it. How can i upgrade without losing my historic data? otherwise did i missed anywhere to update any parameters in the values file? Please advise, thank you in advance!!

stale[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Any further update will cause the issue/pull request to no longer be considered stale. Thank you for your contributions.

stale[bot] commented 1 year ago

This issue is being automatically closed due to inactivity.

holooloo commented 1 year ago

i have the same issue Chart version 44.3.0

gyohuangxin commented 1 year ago

I have the same issue too with chart kube-prometheus-stack-51.0.3.

nskumz commented 1 year ago

does anyone solved this issue??? whats wrong with this kube-prom-stack helm chart. regardless of the version it doesnt showing up the any data in grafana default dahsboards

avnerse3 commented 1 year ago

same here with chart 51.10 clean install on AKS running 1.24.9

rnietoe commented 10 months ago

we want to remove all these alerts rules

based on this article https://grafana.com/docs/grafana-cloud/monitor-infrastructure/kubernetes-monitoring/configuration/configure-infrastructure-manually/helm-operator-migration/import_rules/#disable-local-prometheus-rules-evaluation I tried disabling evaluation of local Prometheus rules.

however, setting defaultRules: create: false causes CPU metrics to be lost. In my case there is no other problem. i.e. I do have memory metrics

it looks like recording rules are needed (https://grafana.com/docs/grafana/latest/alerting/fundamentals/alert-rules/recording-rules/). If you set defaultRules: create: false these alerting rules are not created and you lose metrics.

KOCM0C commented 9 months ago

I have the same issue too chart 56.01 k8s v1.26.5

ShehrozWaseem commented 3 months ago

facing same issue cant see any metrics in dashboard id 315 have everything up and running Kubectl version: client (1.22) and server (1.28) helm info: NAME CHART VERSION APP VERSION
prometheus-community/kube-prometheus-stack 61.7.0 v0.75.2

prometheus-community / helm-charts