Bug 1729529
| Summary: | Inconsistent presentation of memory statistics in Grafana creating confusion | ||
|---|---|---|---|
| Product: | OpenShift Container Platform | Reporter: | Matthew Robson <mrobson> |
| Component: | Monitoring | Assignee: | Pawel Krupa <pkrupa> |
| Status: | CLOSED ERRATA | QA Contact: | Viacheslav <mviaches> |
| Severity: | medium | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 3.11.0 | CC: | alegrand, anpicker, erooth, mloibl, pkrupa, surbania |
| Target Milestone: | --- | ||
| Target Release: | 4.2.0 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | No Doc Update | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2019-10-16 06:29:43 UTC | Type: | Bug |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
I created an issue on the upstream dependency that ships these dashboards, we'll need to discuss with the community what would be best to show consistently. I believe working-set-bytes is the best one to show in aggregations and when drilling down to pods/containers, show the different types of memory metrics. I believe we fixed it with https://github.com/openshift/cluster-monitoring-operator/pull/442 so it should be available in 4.2. Fixed. Checked on:4.2.0-0.nightly-2019-08-23-004712 Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2019:2922 |
Description of problem: The way memory usage is shown in different dashboards is causing some confusion for customers. The first instance is from the 'grafana-dashboard-k8s-resources-cluster' config and uses a sum of container_memory_rss by namespace to show memory in the 'K8s / Compute Resource / Cluster' view: "targets": [ { "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", "legendLink": null, "step": 10 } ], Code: https://github.com/openshift/cluster-monitoring-operator/blob/release-3.11/assets/grafana/dashboard-definitions.yaml#L4033 The first instance is from the grafana-dashboard-k8s-resources-namespace config and used a sum of container_memory_usage_bytes by pod in the 'K8s / Compute Resource / NAmespaces' view: "targets": [ { "expr": "sum(container_memory_usage_bytes{namespace=\"$namespace\", container_name!=\"\"}) by (pod_name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod_name}}", "legendLink": null, "step": 10 } ], Code: https://github.com/openshift/cluster-monitoring-operator/blob/release-3.11/assets/grafana/dashboard-definitions.yaml#L4845 Between these 2 metrics, container_memory_usage_bytes and container_memory_rss are 2 independent metrics generated from cAdvisor: container_memory_usage_bytes is generated from the cgroup memory.usage_in_bytes which includes rss, cache and swap where as container_memory_rss is based on the rss value from memory.stat. Within Grafana, the cluster view memory pane shows "Memory Usage w/o Cache" where as the Namespace pane is showing "Memory Usage", but I don't think that is a clear enough distinction to end users that they're reporting 2 different values from 2 different metric sources. Version-Release number of selected component (if applicable): 3.11 How reproducible: View the resources in grafana Actual results: Memory reporter is different and can confuse end users Expected results: Memory is aligned or we better explain what it is we are reporting on Additional info: