How to obtain usage metrics for OpenShift capacity management
Overview
This article provides examples of the self-service commands you can execute against your UKCloud for Managed OpenShift cluster to gain usage metrics and statistics to help manage your ongoing capacity requirements.
A UKCloud deployed OpenShift cluster comprises a single Starter Pack, plus a number of Runtime Packs as required by your workload demands.
Starter Pack - Provides the foundation OpenShift services to support the orchestration and management of the container runtime environments, along with additional infrastructure nodes to support services such as log aggregation and metric collection.
Each customer environment requires only a single Starter Pack.
Runtime Pack - Nodes on which your containerised applications and services run. You have the choice of difference size Runtime Packs.
You can call off Runtime Packs individually, however, we strongly recommend the initial deployment of two or more Runtime Packs for production environments to provide resilience to your workloads in the event of the failure of a single Runtime Pack.
To add or remove Runtime Packs, raise a Service Request via the My Calls section of the UKCloud Portal.
The sections in this guide, show you how to obtain various statistics about the platform, including:
Overall status of each node in the cluster
CPU usage
Memory usage
Block storage consumption (used by persistent volumes)
Object storage consumption (used by the registry)
Intended audience
To complete the steps in this guide you must have access to and a working knowledge of oc, the OpenStack command-line client (CLI). For more information, see OpenShift's Get Started with the CLI.
Identifying capacity statistics
The OpenStack CLI provides the ability to list the nodes in your cluster and then retrieve status information about a specific node. You can use this information to determine if the nodes in your cluster have enough available disk space and memory.
To obtain the list of nodes that make up the OpenStack cluster, enter the following command:
oc get nodesThe results show you the overall readiness of each node and their role. For example:
$ oc get nodes NAME STATUS ROLES AGE VERSION master-infra-0.5623-f84e8e Ready master 310d v1.11.0+d4cacc0 master-infra-1.5623-f84e8e Ready master 310d v1.11.0+d4cacc0 master-infra-2.5623-f84e8e Ready master 310d v1.11.0+d4cacc0 worker-infra-0.5623-f84e8e Ready infra 310d v1.11.0+d4cacc0 worker-infra-1.5623-f84e8e Ready infra 310d v1.11.0+d4cacc0 worker-tenant-s-0.5623-f84e8e Ready compute 310d v1.11.0+d4cacc0 worker-tenant-s-1.5623-f84e8e Ready compute 310d v1.11.0+d4cacc0You can display a summary of the resource usage on each node using the following command:
oc adm top nodesThe results show the percentage and absolute value of CPU and MEMORY usage on each node. For example:
$ oc adm top nodes NAME CPU(cores) CPU% MEMORY(bytes) MEMORY% master-infra-0.5623-f84e8e 276m 7% 3367Mi 22% master-infra-1.5623-f84e8e 258m 6% 3597Mi 23% master-infra-2.5623-f84e8e 370m 9% 4167Mi 27% worker-infra-0.5623-f84e8e 281m 15% 7697Mi 50% worker-infra-1.5623-f84e8e 566m 31% 7946Mi 52% worker-tenant-s-0.5623-f84e8e 405m 22% 12229Mi 80% worker-tenant-s-1.5623-f84e8e 258m 14% 2451Mi 16%Note
For the
oc adm topcommand to work, Hawkular metrics must be running in your cluster. This command can also report on images and pods.To find the detailed status of a node, enter the following command:
oc describe node <nodename>The results show a large amount of information about disk status (pressure and out-of-disk), memory pressure and usage, CPU usage and resource consumption from any running pods. For example:
$ oc describe node worker-tenant-s-0.5623-f84e8e Name: worker-tenant-s-0.5623-f84e8e Roles: compute Labels: beta.kubernetes.io/arch=amd64 beta.kubernetes.io/instance-type=363e3898-1fb5-4d7f-b0b6-9466fe2448c2 beta.kubernetes.io/os=linux failure-domain.beta.kubernetes.io/region=regionOne failure-domain.beta.kubernetes.io/zone=00021-2 internet=true kubernetes.io/hostname=worker-tenant-s-0.5623-f84e8e logging-infra-fluentd=true node-role.kubernetes.io/compute=true tenant=true Annotations: node.openshift.io/md5sum=5d575cf13ea3d2813ed294fa00e72234 volumes.kubernetes.io/controller-managed-attach-detach=true CreationTimestamp: Mon, 07 Jan 2019 17:50:31 +0000 Taints: <none> Unschedulable: false Conditions: Type Status LastHeartbeatTime LastTransitionTime Reason Message ---- ------ ----------------- ------------------ ------ ------- OutOfDisk False Thu, 14 Nov 2019 10:13:12 +0000 Mon, 04 Nov 2019 14:44:23 +0000 KubeletHasSufficientDisk kubelet has sufficient disk space available MemoryPressure False Thu, 14 Nov 2019 10:13:12 +0000 Mon, 04 Nov 2019 14:44:23 +0000 KubeletHasSufficientMemory kubelet has sufficient memory available DiskPressure False Thu, 14 Nov 2019 10:13:12 +0000 Mon, 04 Nov 2019 14:44:23 +0000 KubeletHasNoDiskPressure kubelet has no disk pressure PIDPressure False Thu, 14 Nov 2019 10:13:12 +0000 Mon, 07 Jan 2019 17:50:31 +0000 KubeletHasSufficientPID kubelet has sufficient PID available Ready True Thu, 14 Nov 2019 10:13:12 +0000 Mon, 04 Nov 2019 14:44:23 +0000 KubeletReady kubelet is posting ready status Addresses: InternalIP: 10.3.1.7 Hostname: worker-tenant-s-0.5623-f84e8e Capacity: cpu: 2 hugepages-1Gi: 0 hugepages-2Mi: 0 memory: 16266524Ki pods: 250 Allocatable: cpu: 1800m hugepages-1Gi: 0 hugepages-2Mi: 0 memory: 15639836Ki pods: 250 System Info: Machine ID: 4c43666486044496beef8fc81fdda75a System UUID: 4C436664-8604-4496-BEEF-8FC81FDDA75A Boot ID: e95bc874-b6d2-4284-9215-e589c91836a2 Kernel Version: 3.10.0-1062.4.1.el7.x86_64 OS Image: Red Hat Enterprise Linux Server 7.7 (Maipo) Operating System: linux Architecture: amd64 Container Runtime Version: docker://1.13.1 Kubelet Version: v1.11.0+d4cacc0 Kube-Proxy Version: v1.11.0+d4cacc0 ProviderID: openstack:///4c436664-8604-4496-beef-8fc81fdda75a Non-terminated Pods: (58 in total) Namespace Name CPU Requests CPU Limits Memory Requests Memory Limits --------- ---- ------------ ---------- --------------- ------------- argo argo-ui-78b65f59b5-s2fgh 0 (0%) 0 (0%) 0 (0%) 0 (0%) argo workflow-controller-685875596-4cttz 0 (0%) 0 (0%) 0 (0%) 0 (0%) backup-test ruby-ex-2-gj927 0 (0%) 0 (0%) 0 (0%) 0 (0%) bluegreen blue-1-nkgr5 0 (0%) 0 (0%) 0 (0%) 0 (0%) bluegreen green-1-h2n6z 0 (0%) 0 (0%) 0 (0%) 0 (0%) connect-test ruby-ex-1-597fv 0 (0%) 0 (0%) 0 (0%) 0 (0%) development myapp-1-4sws5 0 (0%) 0 (0%) 0 (0%) 0 (0%) locust-ingress ruby-ex-1-l7dcc 0 (0%) 0 (0%) 0 (0%) 0 (0%) locust locust-master-1-v6q69 0 (0%) 0 (0%) 0 (0%) 0 (0%) pytest pytest-3-gfdt8 0 (0%) 0 (0%) 0 (0%) 0 (0%) new-backup-test ruby-ex-1-8fk75 0 (0%) 0 (0%) 0 (0%) 0 (0%) openshift-logging logging-fluentd-ktvkm 100m (5%) 0 (0%) 756Mi (4%) 756Mi (4%) openshift-monitoring node-exporter-h6czw 10m (0%) 20m (1%) 20Mi (0%) 40Mi (0%) openshift-node sync-krb68 0 (0%) 0 (0%) 0 (0%) 0 (0%) openshift-sdn ovs-nfbjx 100m (5%) 0 (0%) 300Mi (1%) 0 (0%) openshift-sdn sdn-lsbft 100m (5%) 0 (0%) 200Mi (1%) 0 (0%) postgres postgresql-3-wr7gq 0 (0%) 0 (0%) 512Mi (3%) 512Mi (3%) samplepipeline jenkins-1-862f5 0 (0%) 0 (0%) 512Mi (3%) 512Mi (3%) samplepipeline mongodb-1-jzvgd 0 (0%) 0 (0%) 512Mi (3%) 512Mi (3%) samplepipeline nodejs-mongodb-example-1-74qlc 0 (0%) 0 (0%) 512Mi (3%) 512Mi (3%) testproject cotd-2-cnkvq 0 (0%) 0 (0%) 0 (0%) 0 (0%) ... Allocated resources: (Total limits may be over 100 percent, i.e., overcommitted.) Resource Requests Limits -------- -------- ------ cpu 1060m (58%) 1020m (56%) memory 4220Mi (27%) 3612Mi (23%) Events: <none>You can also obtain some of the above information via the REST API, by using something like:
curl -H "Authorization: Bearer $token" -H 'Accept: application/json' https://ocp.\$clustername:8443/api/v1/nodes
For more information about accessing this data, see How to monitor your OpenShift cluster.
Identifying block storage consumption
For capacity management, it's also useful to know much block storage your cluster is using.
To obtain block storage consumption (persistent volumes) information, enter the following command:
oc get pvThe results will look similar to the following, where we can see there is one item consuming 50 GB of storage.
$ oc get pv NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE pvc-08237e69-cff9-11e9-a9ef-fa163e5c5959 5Gi RWO Delete Bound postgres/postgresql tier2 69d pvc-4da7258e-12a6-11e9-ad9d-fa163e12bba5 22Gi RWO Delete Bound openshift-logging/logging-es-0 tier2 310d pvc-58268285-12a5-11e9-ad4e-fa163e5c5959 50Gi RWO Delete Bound openshift-monitoring/prometheus-k8s-db-prometheus-k8s-0 tier2 310d pvc-88a3dfdc-12a5-11e9-ad4e-fa163e5c5959 50Gi RWO Delete Bound openshift-monitoring/prometheus-k8s-db-prometheus-k8s-1 tier2 310d pvc-ac9597cc-12a5-11e9-ad4e-fa163e5c5959 2Gi RWO Delete Bound openshift-monitoring/alertmanager-main-db-alertmanager-main-0 tier2 310d pvc-bf06155b-12a5-11e9-ad4e-fa163e5c5959 2Gi RWO Delete Bound openshift-monitoring/alertmanager-main-db-alertmanager-main-1 tier2 310d pvc-bf2656a8-12a5-11e9-ad9d-fa163e12bba5 10Gi RWO Delete Bound openshift-infra/metrics-cassandra-1 tier2 310d pvc-e61b15ac-12a5-11e9-ad4e-fa163e5c5959 2Gi RWO Delete Bound openshift-monitoring/alertmanager-main-db-alertmanager-main-2 tier2 310d
Identifying object storage consumption
Object storage is used only for the OpenShift registry, therefore, you can calculate how much object storage your cluster is using by adding up the resources in the registry.
To obtain information about resources in the registry, enter the following two commands:
oc adm top imagestreams oc adm top imagesThe output from these commands will look something like the following:
$ oc adm top images NAME IMAGESTREAMTAG P ARENTS USAGE METADATA STORAGE sha256:d2e75dac02681ed5aded89115b736ba5e83011294686cd6d04780aebffc0ff5d openshift/fuse7-eap-openshift (1.0), openshift/jboss-fuse70-eap-openshift (1.0) < none> <none> no 869.2MiB sha256:ba287c1998c4785d968b30862d0718b664ae755ad6d7eec316db91e4f835b29c backup-test/ruby-25-centos7 (latest), new-backup-test/ruby-25-centos7 (latest) < none> <none> no 193.8MiB sha256:57fa0cb158aa31193908df27fc707afcfdd4bdaf93b3286f5602d5f804e1927f openshift/fuse7-java-openshift (1.0), openshift/jboss-fuse70-java-openshift (1.0) < none> <none> no 156.6MiB sha256:be51ee43b1596078a17756f38a0017e9338c902f9094f1ad677844d165a02d43 openshift/fuse7-karaf-openshift (1.0), openshift/jboss-fuse70-karaf-openshift (1.0) < none> <none> no 156.6MiB sha256:58fe06c9b5cf48bbaea47d18511958f51b6b50b764e6d879b65b4715477d0385 openshift/fuse7-console (1.0), openshift/jboss-fuse70-console (1.0) < none> <none> no 91.88MiB sha256:1f8e264d6fe3a1868e29ba2ad85612ccfc380707dd9f2026f2e5058224dced2c openshift/jboss-decisionserver62-openshift (1.2) ... $ oc adm top imagestreams NAME STORAGE IMAGES LAYERS openshift/jboss-eap64-openshift 5.061GiB 9 48 openshift/fuse7-eap-openshift 3.845GiB 5 29 openshift/jboss-processserver64-openshift 3.167GiB 7 41 openshift/jboss-decisionserver64-openshift 3.167GiB 7 41 openshift/rhpam70-businesscentral-openshift 2.948GiB 3 21 openshift/eap-cd-openshift 2.856GiB 6 30 openshift/jboss-datavirt63-openshift 2.607GiB 5 35 openshift/jboss-eap70-openshift 2.595GiB 5 29 openshift/rhpam70-businesscentral-monitoring-openshift 2.569GiB 3 21 openshift/rhpam70-kieserver-openshift 2.433GiB 3 21 openshift/rhpam70-controller-openshift 2.14GiB 3 21Note
For the
oc adm topcommand to work, Hawkular metrics must be running in your cluster. This command can also report on nodes and pods.The appendix at the end of this guide provides some sample Python code to sum the sizes of resources from these two
oc adm topcommands.
Other monitoring options
You can find a simple example of a custom monitoring application, using the REST API described above, in How to monitor your OpenShift cluster and as a UKCloud blog post.
You can also perform monitoring using Hawkular, which is built into the product. You can access Hawkular graphs and monitoring data from the browser-based console at:
https://ocp.<your-deployment-name>.cna.ukcloud.com:8443
Next steps
Further cluster capacity management options are discussed at:
https://blog.openshift.com/full-cluster-capacity-management-monitoring-openshift
Feedback
If you find a problem with this article, click Improve this Doc to make the change yourself or raise an issue in GitHub. If you have an idea for how we could improve any of our services, send an email to feedback@ukcloud.com.