Monitoring

The following sections contain tips to troubleshoot Harvester Monitoring.

Monitoring is unusable

When the Harvester Dashboard is not showing any monitoring metrics, it can be caused by the following reasons.

Monitoring is unusable due to Pod being stuck in Terminating status

Harvester Monitoring pods are deployed randomly on the cluster Nodes. When the Node hosting the pods accidentally goes down, the related pods may become stuck in the Terminating status rendering the Monitoring unusable from the WebUI.

$ kubectl get pods -n cattle-monitoring-system

NAMESPACE                   NAME                                                     READY   STATUS        RESTARTS   AGE
cattle-monitoring-system    prometheus-rancher-monitoring-prometheus-0               3/3     Terminating   0          3d23h
cattle-monitoring-system    rancher-monitoring-admission-create-fwjn9                0/1     Terminating   0          137m
cattle-monitoring-system    rancher-monitoring-crd-create-9wtzf                      0/1     Terminating   0          137m
cattle-monitoring-system    rancher-monitoring-grafana-d9c56d79b-ph4nz               3/3     Terminating   0          3d23h
cattle-monitoring-system    rancher-monitoring-grafana-d9c56d79b-t24sz               0/3     Init:0/2      0          132m

cattle-monitoring-system    rancher-monitoring-kube-state-metrics-5bc8bb48bd-nbd92   1/1     Running       4          4d1h
...

Monitoring can be recovered using CLI commands to force delete the related pods. The cluster will redeploy new pods to replace them.

# Delete each none-running Pod in namespace cattle-monitoring-system.

$ kubectl delete pod --force -n cattle-monitoring-system prometheus-rancher-monitoring-prometheus-0

 pod "prometheus-rancher-monitoring-prometheus-0" force deleted


$ kubectl delete pod --force -n cattle-monitoring-system rancher-monitoring-admission-create-fwjn9

$ kubectl delete pod --force -n cattle-monitoring-system rancher-monitoring-crd-create-9wtzf

$ kubectl delete pod --force -n cattle-monitoring-system rancher-monitoring-grafana-d9c56d79b-ph4nz 

$ kubectl delete pod --force -n cattle-monitoring-system rancher-monitoring-grafana-d9c56d79b-t24sz

Wait for a few minutes so that the new pods are created and readied for the Monitoring dashboard to be usable again.

$ kubectl get pods -n cattle-monitoring-system 

NAME                                                     READY   STATUS     RESTARTS   AGE
prometheus-rancher-monitoring-prometheus-0               0/3     Init:0/1   0          98s
rancher-monitoring-grafana-d9c56d79b-cp86w               0/3     Init:0/2   0          27s
...

$ kubectl get pods -n cattle-monitoring-system 

NAME                                                     READY   STATUS    RESTARTS   AGE
prometheus-rancher-monitoring-prometheus-0               3/3     Running   0          7m57s
rancher-monitoring-grafana-d9c56d79b-cp86w               3/3     Running   0          6m46s
...

Expand PV/Volume Size

Harvester integrates Longhorn as the default storage provider.

Harvester Monitoring uses Persistent Volume (PV) to store running data. When a cluster has been running for a certain time, the Persistent Volume may need to expand its size.

Based on the Longhorn Volume expansion guide, Harvester illustrates how to expand the volume size.

View Volume

From Embedded Longhorn WebUI

Access the embedded Longhorn WebUI according to this document.

The Longhorn dashboard default view.

Click Volume to list all existing volumes.

From CLI

You can also use kubectl to get all Volumes.

# kubectl get pvc -A
NAMESPACE                  NAME                                                                                             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS         AGE
cattle-monitoring-system   alertmanager-rancher-monitoring-alertmanager-db-alertmanager-rancher-monitoring-alertmanager-0   Bound    pvc-1b2fbbe9-14b1-4a65-941a-7d5645a89977   5Gi        RWO            harvester-longhorn   43h
cattle-monitoring-system   prometheus-rancher-monitoring-prometheus-db-prometheus-rancher-monitoring-prometheus-0           Bound    pvc-7c6dcb61-51a9-4a38-b4c5-acaa11788978   50Gi       RWO            harvester-longhorn   43h
cattle-monitoring-system   rancher-monitoring-grafana                                                                       Bound    pvc-b2b2c07c-f7cd-4965-90e6-ac3319597bf7   2Gi        RWO            harvester-longhorn   43h

# kubectl get volume -A
NAMESPACE         NAME                                       STATE      ROBUSTNESS   SCHEDULED   SIZE          NODE     AGE
longhorn-system   pvc-1b2fbbe9-14b1-4a65-941a-7d5645a89977   attached   degraded                 5368709120    harv31   43h
longhorn-system   pvc-7c6dcb61-51a9-4a38-b4c5-acaa11788978   attached   degraded                 53687091200   harv31   43h
longhorn-system   pvc-b2b2c07c-f7cd-4965-90e6-ac3319597bf7   attached   degraded                 2147483648    harv31   43h

Scale Down a Deployment

To detach the Volume, you need to scale down the deployment that uses the Volume.

The example below is against the PVC claimed by rancher-monitoring-grafana.

Find the deployment in the namespace cattle-monitoring-system.

# kubectl get deployment -n cattle-monitoring-system
NAME                                    READY   UP-TO-DATE   AVAILABLE   AGE
rancher-monitoring-grafana              1/1     1            1           43h  // target deployment
rancher-monitoring-kube-state-metrics   1/1     1            1           43h
rancher-monitoring-operator             1/1     1            1           43h
rancher-monitoring-prometheus-adapter   1/1     1            1           43h

Scale down the deployment rancher-monitoring-grafana to 0.

# kubectl scale --replicas=0 deployment/rancher-monitoring-grafana -n cattle-monitoring-system

Check the deployment and the volume.

# kubectl get deployment -n cattle-monitoring-system
NAME                                    READY   UP-TO-DATE   AVAILABLE   AGE
rancher-monitoring-grafana              0/0     0            0           43h  // scaled down
rancher-monitoring-kube-state-metrics   1/1     1            1           43h
rancher-monitoring-operator             1/1     1            1           43h
rancher-monitoring-prometheus-adapter   1/1     1            1           43h

# kubectl get volume -A
NAMESPACE         NAME                                       STATE      ROBUSTNESS   SCHEDULED   SIZE          NODE     AGE
longhorn-system   pvc-1b2fbbe9-14b1-4a65-941a-7d5645a89977   attached   degraded                 5368709120    harv31   43h
longhorn-system   pvc-7c6dcb61-51a9-4a38-b4c5-acaa11788978   attached   degraded                 53687091200   harv31   43h
longhorn-system   pvc-b2b2c07c-f7cd-4965-90e6-ac3319597bf7   detached   unknown                  2147483648             43h  // volume is detached

Expand Volume

In the Longhorn WebUI, the related volume becomes Detached. Click the icon in the Operation column, and select Expand Volume.

Input a new size, and Longhorn will expand the volume to this size.

Scale Up a Deployment

After the Volume is expanded to target size, you need to scale up the aforementioned deployment to its original replicas. For the above example of rancher-monitoring-grafana, the original replicas is 1.

# kubectl scale --replicas=1 deployment/rancher-monitoring-grafana -n cattle-monitoring-system

Check the deployment again.

# kubectl get deployment -n cattle-monitoring-system
NAME                                    READY   UP-TO-DATE   AVAILABLE   AGE
rancher-monitoring-grafana              1/1     1            1           43h  // scaled up
rancher-monitoring-kube-state-metrics   1/1     1            1           43h
rancher-monitoring-operator             1/1     1            1           43h
rancher-monitoring-prometheus-adapter   1/1     1            1           43h

The Volume is attached to the new POD.

To now, the Volume is expanded to the new size and the POD is using it smoothly.

Fail to Enable rancher-monitoring Addon

You may encounter this when you install the Harvester v1.3.0 or higher version cluster with the minimal 250 GB disk per hardware requirements.

Reproduce Steps

1. Install the Harvester v1.3.0 cluster.

2. Enable the rancher-monitoring addon, you will observe:

• The POD prometheus-rancher-monitoring-prometheus-0 in cattle-monitoring-system namespace fails to start due to PVC attached failed.

  $ kubectl get pods -n cattle-monitoring-system
  NAME                                                     READY   STATUS      RESTARTS   AGE
  alertmanager-rancher-monitoring-alertmanager-0           2/2     Running     0          3m22s
  helm-install-rancher-monitoring-4b5mx                    0/1     Completed   0          3m41s
  prometheus-rancher-monitoring-prometheus-0               0/3     Init:0/1    0          3m21s // stuck in this status
  rancher-monitoring-grafana-d6f466988-hgpkb               4/4     Running     0          3m26s
  rancher-monitoring-kube-state-metrics-7659b76cc4-66sr7   1/1     Running     0          3m26s
  rancher-monitoring-operator-595476bc84-7hdxj             1/1     Running     0          3m25s
  rancher-monitoring-prometheus-adapter-55dc9ccd5d-pcrpk   1/1     Running     0          3m26s
  rancher-monitoring-prometheus-node-exporter-pbzv4        1/1     Running     0          3m26s
  
  $ kubectl describe pod -n cattle-monitoring-system prometheus-rancher-monitoring-prometheus-0
  Name:             prometheus-rancher-monitoring-prometheus-0
  Namespace:        cattle-monitoring-system
  Priority:         0
  Service Account:  rancher-monitoring-prometheus
  ...
  Events:
    Type     Reason              Age                    From                     Message
    ----     ------              ----                   ----                     -------
    Warning  FailedScheduling    3m48s (x3 over 4m15s)  default-scheduler        0/1 nodes are available: pod has unbound immediate PersistentVolumeClaims. preemption: 0/1 nodes are available: 1 Preemption is not helpful for scheduling..
    Normal   Scheduled           3m44s                  default-scheduler        Successfully assigned cattle-monitoring-system/prometheus-rancher-monitoring-prometheus-0 to harv41
    Warning  FailedMount         101s                   kubelet                  Unable to attach or mount volumes: unmounted volumes=[prometheus-rancher-monitoring-prometheus-db], unattached volumes=[prometheus-rancher-monitoring-prometheus-db], failed to process volumes=[]: timed out waiting for the condition
    Warning  FailedAttachVolume  90s (x9 over 3m42s)    attachdetach-controller  AttachVolume.Attach failed for volume "pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0" : rpc error: code = Aborted desc = volume pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0 is not ready for workloads
  
  $ kubectl get pvc -A
  NAMESPACE                  NAME                                                                                             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS           AGE
  cattle-monitoring-system   prometheus-rancher-monitoring-prometheus-db-prometheus-rancher-monitoring-prometheus-0           Bound    pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0   50Gi       RWO            harvester-longhorn     7m12s
  
  $ kubectl get volume -A
  NAMESPACE         NAME                                       DATA ENGINE   STATE      ROBUSTNESS   SCHEDULED   SIZE          NODE     AGE
  longhorn-system   pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0   v1            detached   unknown                  53687091200            6m55s
  

• The Longhorn manager is unable to schedule the replica.

  $ kubectl logs -n longhorn-system longhorn-manager-bf65b | grep "pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0"
  
  time="2024-02-19T10:12:56Z" level=error msg="There's no available disk for replica pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0-r-dcb129fd, size 53687091200" func="schedule
  r.(*ReplicaScheduler).ScheduleReplica" file="replica_scheduler.go:95"
  time="2024-02-19T10:12:56Z" level=warning msg="Failed to schedule replica" func="controller.(*VolumeController).reconcileVolumeCondition" file="volume_controller.go:169
  4" accessMode=rwo controller=longhorn-volume frontend=blockdev migratable=false node=harv41 owner=harv41 replica=pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0-r-dcb129fd sta
  te= volume=pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0
  ...

Workaround

1. Disable the rancher-monitoring addon if you have alreay enabled it.

   All pods in cattle-monitoring-system are deleted but the PVCs are retained. For more information, see [Addons].

   $ kubectl get pods -n cattle-monitoring-system
   No resources found in cattle-monitoring-system namespace.
   
   $ kubectl get pvc -n cattle-monitoring-system
   NAME                                                                                             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS         AGE
   alertmanager-rancher-monitoring-alertmanager-db-alertmanager-rancher-monitoring-alertmanager-0   Bound    pvc-cea6316e-f74f-4771-870b-49edb5442819   5Gi        RWO            harvester-longhorn   14m
   prometheus-rancher-monitoring-prometheus-db-prometheus-rancher-monitoring-prometheus-0           Bound    pvc-bbe8760d-926c-484a-851c-b8ec29ae05c0   50Gi       RWO            harvester-longhorn   14m

2. Delete the PVC named prometheus, but retain the PVC named alertmanager.

   $ kubectl delete pvc -n cattle-monitoring-system prometheus-rancher-monitoring-prometheus-db-prometheus-rancher-monitoring-prometheus-0
   persistentvolumeclaim "prometheus-rancher-monitoring-prometheus-db-prometheus-rancher-monitoring-prometheus-0" deleted
   
   $ kubectl get pvc -n cattle-monitoring-system
   NAME                                                                                             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS         AGE
   alertmanager-rancher-monitoring-alertmanager-db-alertmanager-rancher-monitoring-alertmanager-0   Bound    pvc-cea6316e-f74f-4771-870b-49edb5442819   5Gi        RWO            harvester-longhorn   16m

3. On the Addons screen of the Harvester UI, select ⋮ (menu icon) and then select Edit YAML.

   

4. As indicated below, change the two occurrences of the number 50 to 30 under prometheusSpec, and then save. The prometheus feature will use a 30GiB disk to store data.

   

   Alternatively, you can use kubectl to edit the object.

   kubectl edit addons.harvesterhci.io -n cattle-monitoring-system rancher-monitoring

           retentionSize: 50GiB // Change 50 to 30
           storageSpec:
             volumeClaimTemplate:
               spec:
                 accessModes:
                   - ReadWriteOnce
                 resources:
                   requests:
                     storage: 50Gi // Change 50 to 30
                 storageClassName: harvester-longhorn

5. Enable the rancher-monitoring addon and wait for a few minutes..

6. All pods are successfully deployed, and the rancher-monitoring feature is available.

   $ kubectl get pods -n cattle-monitoring-system
   NAME                                                     READY   STATUS      RESTARTS   AGE
   alertmanager-rancher-monitoring-alertmanager-0           2/2     Running     0          3m52s
   helm-install-rancher-monitoring-s55tq                    0/1     Completed   0          4m17s
   prometheus-rancher-monitoring-prometheus-0               3/3     Running     0          3m51s
   rancher-monitoring-grafana-d6f466988-hkv6f               4/4     Running     0          3m55s
   rancher-monitoring-kube-state-metrics-7659b76cc4-ght8x   1/1     Running     0          3m55s
   rancher-monitoring-operator-595476bc84-r96bp             1/1     Running     0          3m55s
   rancher-monitoring-prometheus-adapter-55dc9ccd5d-vtssc   1/1     Running     0          3m55s
   rancher-monitoring-prometheus-node-exporter-lgb88        1/1     Running     0          3m55s