This is a cache of https://docs.okd.io/4.6/scalability_and_performance/recommended-host-practices.html. It is a snapshot of the page at 2024-11-22T02:45:10.431+0000.
Recommended host practices | Scalability and performance | OKD 4.6
  1. Documentation
×

Recommended host practices

The OKD node configuration file contains important options. For example, two parameters control the maximum number of pods that can be scheduled to a node: podsPerCore and maxPods.

When both options are in use, the lower of the two values limits the number of pods on a node. Exceeding these values can result in:

  • Increased CPU utilization.

  • Slow pod scheduling.

  • Potential out-of-memory scenarios, depending on the amount of memory in the node.

  • Exhausting the pool of IP addresses.

  • Resource overcommitting, leading to poor user application performance.

In Kubernetes, a pod that is holding a single container actually uses two containers. The second container is used to set up networking prior to the actual container starting. Therefore, a system running 10 pods will actually have 20 containers running.

podsPerCore sets the number of pods the node can run based on the number of processor cores on the node. For example, if podsPerCore is set to 10 on a node with 4 processor cores, the maximum number of pods allowed on the node will be 40.

kubeletConfig:
  podsPerCore: 10

Setting podsPerCore to 0 disables this limit. The default is 0. podsPerCore cannot exceed maxPods.

maxPods sets the number of pods the node can run to a fixed value, regardless of the properties of the node.

 kubeletConfig:
    maxPods: 250

The kubelet configuration is currently serialized as an Ignition configuration, so it can be directly edited. However, there is also a new kubelet-config-controller added to the Machine Config Controller (MCC). This allows you to create a KubeletConfig custom resource (CR) to edit the kubelet parameters.

As the fields in the kubeletConfig object are passed directly to the kubelet from upstream Kubernetes, the kubelet validates those values directly. Invalid values in the kubeletConfig object might cause cluster nodes to become unavailable. For valid values, see the Kubernetes documentation.
Procedure

  1. View the available machine configuration objects that you can select:

    $ oc get machineconfig

    By default, the two kubelet-related configs are 01-master-kubelet and 01-worker-kubelet.

  2. To check the current value of max pods per node, run:

    # oc describe node <node-ip> | grep Allocatable -A6

    Look for value: pods: <value>.

    For example:

    # oc describe node ip-172-31-128-158.us-east-2.compute.internal | grep Allocatable -A6
    Example output
    Allocatable:
 attachable-volumes-aws-ebs:  25
 cpu:                         3500m
 hugepages-1Gi:               0
 hugepages-2Mi:               0
 memory:                      15341844Ki
 pods:                        250

  3. To set the max pods per node on the worker nodes, create a custom resource file that contains the kubelet configuration. For example, change-maxPods-cr.yaml:

    apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: set-max-pods
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: large-pods
  kubeletConfig:
    maxPods: 500

    The rate at which the kubelet talks to the API server depends on queries per second (QPS) and burst values. The default values, 50 for kubeAPIQPS and 100 for kubeAPIBurst, are good enough if there are limited pods running on each node. Updating the kubelet QPS and burst rates is recommended if there are enough CPU and memory resources on the node:

    apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: set-max-pods
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: large-pods
  kubeletConfig:
    maxPods: <pod_count>
    kubeAPIBurst: <burst_rate>
    kubeAPIQPS: <QPS>

    1. Update the machine config pool for workers with the label:

      $ oc label machineconfigpool worker custom-kubelet=large-pods

    2. Create the KubeletConfig object:

      $ oc create -f change-maxPods-cr.yaml

    3. Verify that the KubeletConfig object is created:

      $ oc get kubeletconfig

      This should return set-max-pods.

      Depending on the number of worker nodes in the cluster, wait for the worker nodes to be rebooted one by one. For a cluster with 3 worker nodes, this could take about 10 to 15 minutes.

  4. Check for maxPods changing for the worker nodes:

    $ oc describe node

    1. Verify the change by running:

      $ oc get kubeletconfigs set-max-pods -o yaml

      This should show a status of True and type:Success

Procedure

By default, only one machine is allowed to be unavailable when applying the kubelet-related configuration to the available worker nodes. For a large cluster, it can take a long time for the configuration change to be reflected. At any time, you can adjust the number of machines that are updating to speed up the process.

  1. Edit the worker machine config pool:

    $ oc edit machineconfigpool worker

  2. Set maxUnavailable to the desired value.

    spec:
  maxUnavailable: <node_count>

    When setting the value, consider the number of worker nodes that can be unavailable without affecting the applications running on the cluster.

The control plane node resource requirements depend on the number of nodes in the cluster. The following control plane node size recommendations are based on the results of control plane density focused testing. The control plane tests create the following objects across the cluster in each of the namespaces depending on the node counts:

  • 12 image streams

  • 3 build configurations

  • 6 builds

  • 1 deployment with 2 pod replicas mounting two secrets each

  • 2 deployments with 1 pod replica mounting two secrets

  • 3 services pointing to the previous deployments

  • 3 routes pointing to the previous deployments

  • 10 secrets, 2 of which are mounted by the previous deployments

  • 10 config maps, 2 of which are mounted by the previous deployments

Number of worker nodes Cluster load (namespaces) CPU cores Memory (GB)

25

500

4

16

100

1000

8

32

250

4000

16

96

On a large and dense cluster with three masters or control plane nodes, the CPU and memory usage will spike up when one of the nodes is stopped, rebooted or fails. The failures can be due to unexpected issues with power, network or underlying infrastructure in addition to intentional cases where the cluster is restarted after shutting it down to save costs. The remaining two control plane nodes must handle the load in order to be highly available which leads to increase in the resource usage. This is also expected during upgrades because the masters are cordoned, drained, and rebooted serially to apply the operating system updates, as well as the control plane Operators update. To avoid cascading failures, keep the overall CPU and memory resource usage on the control plane nodes to at most 60% of all available capacity to handle the resource usage spikes. Increase the CPU and memory on the control plane nodes accordingly to avoid potential downtime due to lack of resources.

The node sizing varies depending on the number of nodes and object counts in the cluster. It also depends on whether the objects are actively being created on the cluster. During object creation, the control plane is more active in terms of resource usage compared to when the objects are in the running phase.

If you used an installer-provisioned infrastructure installation method, you cannot modify the control plane node size in a running OKD 4.6 cluster. Instead, you must estimate your total node count and use the suggested control plane node size during installation.

The recommendations are based on the data points captured on OKD clusters with OpenShiftSDN as the network plug-in.

In OKD 4.6, half of a CPU core (500 millicore) is now reserved by the system by default compared to OKD 3.11 and previous versions. The sizes are determined taking that into consideration.

When you have overloaded AWS master nodes in a cluster and the master nodes require more resources, you can increase the flavor size of the master instances.

It is recommended to backup etcd before increasing the flavor size of the AWS master instances.
Prerequisites

  • You have an IPI (installer-provisioned infrastructure) or UPI (user-provisioned infrastructure) cluster on AWS.

Procedure

  1. Open the AWS console, fetch the master instances.

  2. Stop one master instance.

  3. Select the stopped instance, and click Actions → Instance Settings → Change instance type.

  4. Change the instance to a larger type, ensuring that the type is the same base as the previous selection, and apply changes. For example, you can change m5.xlarge to m5.2xlarge or m5.4xlarge.

  5. Backup the instance, and repeat the steps for the next master instance.

Additional resources

For large and dense clusters, etcd can suffer from poor performance if the keyspace grows too large and exceeds the space quota. Periodically maintain and defragment etcd to free up space in the data store. Monitor Prometheus for etcd metrics and defragment it when required; otherwise, etcd can raise a cluster-wide alarm that puts the cluster into a maintenance mode that accepts only key reads and deletes.

Monitor these key metrics:

  • etcd_server_quota_backend_bytes, which is the current quota limit

  • etcd_mvcc_db_total_size_in_use_in_bytes, which indicates the actual database usage after a history compaction

  • etcd_debugging_mvcc_db_total_size_in_bytes, which shows the database size, including free space waiting for defragmentation

For more information about defragmenting etcd, see the "Defragmenting etcd data" section.

Because etcd writes data to disk and persists proposals on disk, its performance depends on disk performance. Slow disks and disk activity from other processes can cause long fsync latencies. Those latencies can cause etcd to miss heartbeats, not commit new proposals to the disk on time, and ultimately experience request timeouts and temporary leader loss. Run etcd on machines that are backed by SSD or NVMe disks with low latency and high throughput. Consider single-level cell (SLC) solid-state drives (SSDs), which provide 1 bit per memory cell, are durable and reliable, and are ideal for write-intensive workloads.

Some key metrics to monitor on a deployed OKD cluster are p99 of etcd disk write ahead log duration and the number of etcd leader changes. Use Prometheus to track these metrics.

  • The etcd_disk_wal_fsync_duration_seconds_bucket metric reports the etcd disk fsync duration.

  • The etcd_server_leader_changes_seen_total metric reports the leader changes.

  • To rule out a slow disk and confirm that the disk is reasonably fast, verify that the 99th percentile of the etcd_disk_wal_fsync_duration_seconds_bucket is less than 10 ms.

To validate the hardware for etcd before or after you create the OKD cluster, you can use an I/O benchmarking tool called fio.

Prerequisites

  • Container runtimes such as Podman or Docker are installed on the machine that you’re testing.

  • Data is written to the /var/lib/etcd path.

Procedure

  • Run fio and analyze the results:

    • If you use Podman, run this command:

      $ sudo podman run --volume /var/lib/etcd:/var/lib/etcd:Z quay.io/openshift-scale/etcd-perf

    • If you use Docker, run this command:

      $ sudo docker run --volume /var/lib/etcd:/var/lib/etcd:Z quay.io/openshift-scale/etcd-perf

The output reports whether the disk is fast enough to host etcd by comparing the 99th percentile of the fsync metric captured from the run to see if it is less than 10 ms.

Because etcd replicates the requests among all the members, its performance strongly depends on network input/output (I/O) latency. High network latencies result in etcd heartbeats taking longer than the election timeout, which results in leader elections that are disruptive to the cluster. A key metric to monitor on a deployed OKD cluster is the 99th percentile of etcd network peer latency on each etcd cluster member. Use Prometheus to track the metric.

The histogram_quantile(0.99, rate(etcd_network_peer_round_trip_time_seconds_bucket[2m])) metric reports the round trip time for etcd to finish replicating the client requests between the members. Ensure that it is less than 50 ms.

Manual defragmentation must be performed periodically to reclaim disk space after etcd history compaction and other events cause disk fragmentation.

History compaction is performed automatically every five minutes and leaves gaps in the back-end database. This fragmented space is available for use by etcd, but is not available to the host file system. You must defragment etcd to make this space available to the host file system.

Because etcd writes data to disk, its performance strongly depends on disk performance. Consider defragmenting etcd every month, twice a month, or as needed for your cluster. You can also monitor the etcd_db_total_size_in_bytes metric to determine whether defragmentation is necessary.

Defragmenting etcd is a blocking action. The etcd member will not response until defragmentation is complete. For this reason, wait at least one minute between defragmentation actions on each of the pods to allow the cluster to recover.

Follow this procedure to defragment etcd data on each etcd member.

Prerequisites

  • You have access to the cluster as a user with the cluster-admin role.

Procedure

  1. Determine which etcd member is the leader, because the leader should be defragmented last.

    1. Get the list of etcd pods:

      $ oc get pods -n openshift-etcd -o wide | grep -v quorum-guard | grep etcd
      Example output
      etcd-ip-10-0-159-225.example.redhat.com                3/3     Running     0          175m   10.0.159.225   ip-10-0-159-225.example.redhat.com   <none>           <none>
etcd-ip-10-0-191-37.example.redhat.com                 3/3     Running     0          173m   10.0.191.37    ip-10-0-191-37.example.redhat.com    <none>           <none>
etcd-ip-10-0-199-170.example.redhat.com                3/3     Running     0          176m   10.0.199.170   ip-10-0-199-170.example.redhat.com   <none>           <none>

    2. Choose a pod and run the following command to determine which etcd member is the leader:

      $ oc rsh -n openshift-etcd etcd-ip-10-0-159-225.example.redhat.com etcdctl endpoint status --cluster -w table
      Example output
      Defaulting container name to etcdctl.
Use 'oc describe pod/etcd-ip-10-0-159-225.example.redhat.com -n openshift-etcd' to see all of the containers in this pod.
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
|         ENDPOINT          |        ID        | VERSION | DB SIZE | IS LEADER | IS LEARNER | RAFT TERM | RAFT INDEX | RAFT APPLIED INDEX | ERRORS |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
|  https://10.0.191.37:2379 | 251cd44483d811c3 |   3.4.9 |  104 MB |     false |      false |         7 |      91624 |              91624 |        |
| https://10.0.159.225:2379 | 264c7c58ecbdabee |   3.4.9 |  104 MB |     false |      false |         7 |      91624 |              91624 |        |
| https://10.0.199.170:2379 | 9ac311f93915cc79 |   3.4.9 |  104 MB |      true |      false |         7 |      91624 |              91624 |        |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+

      Based on the IS LEADER column of this output, the https://10.0.199.170:2379 endpoint is the leader. Matching this endpoint with the output of the previous step, the pod name of the leader is etcd-ip-10-0-199-170.example.redhat.com.

  2. Defragment an etcd member.

    1. Connect to the running etcd container, passing in the name of a pod that is not the leader:

      $ oc rsh -n openshift-etcd etcd-ip-10-0-159-225.example.redhat.com

    2. Unset the etcdCTL_ENDPOINTS environment variable:

      sh-4.4# unset etcdCTL_ENDPOINTS

    3. Defragment the etcd member:

      sh-4.4# etcdctl --command-timeout=30s --endpoints=https://localhost:2379 defrag
      Example output
      Finished defragmenting etcd member[https://localhost:2379]

      If a timeout error occurs, increase the value for --command-timeout until the command succeeds.

    4. Verify that the database size was reduced:

      sh-4.4# etcdctl endpoint status -w table --cluster
      Example output
      +---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
|         ENDPOINT          |        ID        | VERSION | DB SIZE | IS LEADER | IS LEARNER | RAFT TERM | RAFT INDEX | RAFT APPLIED INDEX | ERRORS |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+
|  https://10.0.191.37:2379 | 251cd44483d811c3 |   3.4.9 |  104 MB |     false |      false |         7 |      91624 |              91624 |        |
| https://10.0.159.225:2379 | 264c7c58ecbdabee |   3.4.9 |   41 MB |     false |      false |         7 |      91624 |              91624 |        | (1)
| https://10.0.199.170:2379 | 9ac311f93915cc79 |   3.4.9 |  104 MB |      true |      false |         7 |      91624 |              91624 |        |
+---------------------------+------------------+---------+---------+-----------+------------+-----------+------------+--------------------+--------+

      This example shows that the database size for this etcd member is now 41 MB as opposed to the starting size of 104 MB.

    5. Repeat these steps to connect to each of the other etcd members and defragment them. Always defragment the leader last.

      Wait at least one minute between defragmentation actions to allow the etcd pod to recover. Until the etcd pod recovers, the etcd member will not respond.

  3. If any NOSPACE alarms were triggered due to the space quota being exceeded, clear them.

    1. Check if there are any NOSPACE alarms:

      sh-4.4# etcdctl alarm list
      Example output
      memberID:12345678912345678912 alarm:NOSPACE

    2. Clear the alarms:

      sh-4.4# etcdctl alarm disarm

The following infrastructure workloads do not incur OKD worker subscriptions:

  • Kubernetes and OKD control plane services that run on masters

  • The default router

  • The integrated container image registry

  • The HAProxy-based Ingress Controller

  • The cluster metrics collection, or monitoring service, including components for monitoring user-defined projects

  • Cluster aggregated logging

  • Service brokers

  • Red Hat Quay

  • Red Hat OpenShift Container Storage

  • Red Hat Advanced Cluster Manager

  • Red Hat Advanced Cluster Security for Kubernetes

  • Red Hat OpenShift GitOps

  • Red Hat OpenShift Pipelines

Any node that runs any other container, pod, or component is a worker node that your subscription must cover.

Additional resources

By default, the Prometheus Cluster Monitoring stack, which contains Prometheus, Grafana, and AlertManager, is deployed to provide cluster monitoring. It is managed by the Cluster Monitoring Operator. To move its components to different machines, you create and apply a custom config map.

Procedure

  1. Save the following ConfigMap definition as the cluster-monitoring-configmap.yaml file:

    apiVersion: v1
kind: ConfigMap
metadata:
  name: cluster-monitoring-config
  namespace: openshift-monitoring
data:
  config.yaml: |+
    alertmanagerMain:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    prometheusK8s:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    prometheusOperator:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    grafana:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    k8sPrometheusAdapter:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    kubeStateMetrics:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    telemeterClient:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    openshiftStateMetrics:
      nodeSelector:
        node-role.kubernetes.io/infra: ""
    thanosQuerier:
      nodeSelector:
        node-role.kubernetes.io/infra: ""

    Running this config map forces the components of the monitoring stack to redeploy to infrastructure nodes.

  2. Apply the new config map:

    $ oc create -f cluster-monitoring-configmap.yaml

  3. Watch the monitoring pods move to the new machines:

    $ watch 'oc get pod -n openshift-monitoring -o wide'

  4. If a component has not moved to the infra node, delete the pod with this component:

    $ oc delete pod -n openshift-monitoring <pod>

    The component from the deleted pod is re-created on the infra node.

You configure the registry Operator to deploy its pods to different nodes.

Prerequisites

  • Configure additional machine sets in your OKD cluster.

Procedure

  1. View the config/instance object:

    $ oc get configs.imageregistry.operator.openshift.io/cluster -o yaml
    Example output
    apiVersion: imageregistry.operator.openshift.io/v1
kind: Config
metadata:
  creationTimestamp: 2019-02-05T13:52:05Z
  finalizers:
  - imageregistry.operator.openshift.io/finalizer
  generation: 1
  name: cluster
  resourceVersion: "56174"
  selfLink: /apis/imageregistry.operator.openshift.io/v1/configs/cluster
  uid: 36fd3724-294d-11e9-a524-12ffeee2931b
spec:
  httpSecret: d9a012ccd117b1e6616ceccb2c3bb66a5fed1b5e481623
  logging: 2
  managementState: Managed
  proxy: {}
  replicas: 1
  requests:
    read: {}
    write: {}
  storage:
    s3:
      bucket: image-registry-us-east-1-c92e88cad85b48ec8b312344dff03c82-392c
      region: us-east-1
status:
...

  2. Edit the config/instance object:

    $ oc edit configs.imageregistry.operator.openshift.io/cluster

  3. Modify the spec section of the object to resemble the following YAML:

    spec:
  affinity:
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - podAffinityTerm:
          namespaces:
          - openshift-image-registry
          topologyKey: kubernetes.io/hostname
        weight: 100
  logLevel: Normal
  managementState: Managed
  nodeSelector:
    node-role.kubernetes.io/infra: ""

  4. Verify the registry pod has been moved to the infrastructure node.

    1. Run the following command to identify the node where the registry pod is located:

      $ oc get pods -o wide -n openshift-image-registry

    2. Confirm the node has the label you specified:

      $ oc describe node <node_name>

      Review the command output and confirm that node-role.kubernetes.io/infra is in the LABELS list.

You can deploy the router pod to a different machine set. By default, the pod is deployed to a worker node.

Prerequisites

  • Configure additional machine sets in your OKD cluster.

Procedure

  1. View the IngressController custom resource for the router Operator:

    $ oc get ingresscontroller default -n openshift-ingress-operator -o yaml

    The command output resembles the following text:

    apiVersion: operator.openshift.io/v1
kind: IngressController
metadata:
  creationTimestamp: 2019-04-18T12:35:39Z
  finalizers:
  - ingresscontroller.operator.openshift.io/finalizer-ingresscontroller
  generation: 1
  name: default
  namespace: openshift-ingress-operator
  resourceVersion: "11341"
  selfLink: /apis/operator.openshift.io/v1/namespaces/openshift-ingress-operator/ingresscontrollers/default
  uid: 79509e05-61d6-11e9-bc55-02ce4781844a
spec: {}
status:
  availableReplicas: 2
  conditions:
  - lastTransitionTime: 2019-04-18T12:36:15Z
    status: "True"
    type: Available
  domain: apps.<cluster>.example.com
  endpointPublishingStrategy:
    type: LoadBalancerService
  selector: ingresscontroller.operator.openshift.io/deployment-ingresscontroller=default

  2. Edit the ingresscontroller resource and change the nodeSelector to use the infra label:

    $ oc edit ingresscontroller default -n openshift-ingress-operator

    Add the nodeSelector stanza that references the infra label to the spec section, as shown:

      spec:
    nodePlacement:
      nodeSelector:
        matchLabels:
          node-role.kubernetes.io/infra: ""

  3. Confirm that the router pod is running on the infra node.

    1. View the list of router pods and note the node name of the running pod:

      $ oc get pod -n openshift-ingress -o wide
      Example output
      NAME                              READY     STATUS        RESTARTS   AGE       IP           NODE                           NOMINATED NODE   READINESS GATES
router-default-86798b4b5d-bdlvd   1/1      Running       0          28s       10.130.2.4   ip-10-0-217-226.ec2.internal   <none>           <none>
router-default-955d875f4-255g8    0/1      Terminating   0          19h       10.129.2.4   ip-10-0-148-172.ec2.internal   <none>           <none>

      In this example, the running pod is on the ip-10-0-217-226.ec2.internal node.

    2. View the node status of the running pod:

      $ oc get node <node_name> (1)
      1 Specify the <node_name> that you obtained from the pod list.
      Example output
      NAME                          STATUS  ROLES         AGE   VERSION
ip-10-0-217-226.ec2.internal  Ready   infra,worker  17h   v1.19.0

      Because the role list includes infra, the pod is running on the correct node.

The infrastructure node resource requirements depend on the cluster age, nodes, and objects in the cluster, as these factors can lead to an increase in the number of metrics or time series in Prometheus. The following infrastructure node size recommendations are based on the results of cluster maximums and control plane density focused testing.

The sizing recommendations below are only applicable for the Prometheus, Router, and Registry infrastructure components, which are installed during cluster installation. Logging is a day-two operation and its sizing recommendation are given in the last two columns.
Number of worker nodes CPU cores Memory (GB) CPU cores with Logging Memory (GB) with Logging

25

4

16

4

64

100

8

32

8

128

250

16

128

16

128

500

32

128

32

192

These sizing recommendations are based on scale tests, which create a large number of objects across the cluster. These tests include reaching some of the cluster maximums. In the case of 250 and 500 node counts on a OKD 4.6 cluster, these maximums are 10000 namespaces with 61000 pods, 10000 deployments, 181000 secrets, 400 config maps, and so on. Prometheus is a highly memory intensive application; the resource usage depends on various factors including the number of nodes, objects, the Prometheus metrics scraping interval, metrics or time series, and the age of the cluster. The disk size also depends on the retention period. You must take these factors into consideration and size them accordingly.

In OKD 4.6, half of a CPU core (500 millicore) is now reserved by the system by default compared to OKD 3.11 and previous versions. This influences the stated sizing recommendations.

Additional resources