Bug 1991739
| Summary: | WMCO ignores the `Deleting` phase notification event | ||
|---|---|---|---|
| Product: | OpenShift Container Platform | Reporter: | jvaldes |
| Component: | Windows Containers | Assignee: | jvaldes |
| Status: | CLOSED ERRATA | QA Contact: | gaoshang <sgao> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 4.9 | CC: | aos-bugs, aravindh, rrasouli, team-winc |
| Target Milestone: | --- | ||
| Target Release: | 4.9.0 | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | If docs needed, set a value | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2021-10-28 17:41:17 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: | |||
| Bug Depends On: | |||
| Bug Blocks: | 1995341 | ||
@jvaldes, The steps to reproduce seems to indicate that this happens consistently. However you say this behavior is only seen occasionally? How can this be reproduced consistently? Indeed, this behavior happens occasionally. Ran several tests using from 1 to 3 replicas in the machineSet. Sometimes, the `windows-exporter` endpoint is not getting updated by WMCO, removing the information of the deleted machines.
> The steps to reproduce seems to indicate that this happens consistently
Open to suggestions to make that clear.
Marking as VERIFIED to allow the release-4.7/4.8 PRs to merge. Will move this back to ON_QA once that PR merges. Setting status back to ON_QA. oc describoc get node -l kubernetes.io/os=windows -owide
NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
winworker-g8fgw Ready worker 113m v1.22.0-rc.0.1611+9b1230e88478e6 172.31.249.177 172.31.249.177 Windows Server Standard 10.0.19041.508 docker://20.10.5
winworker-pm4kf Ready worker 107m v1.22.0-rc.0.1611+9b1230e88478e6 172.31.249.147 172.31.249.147 Windows Server Standard 10.0.19041.508 docker://20.10.5
[cloud-user@PSI-VM ~/windows-machine-config-operator]> oc describe endpoints -n openshift-windows-machine-config-operator
Name: windows-exporter
Namespace: openshift-windows-machine-config-operator
Labels: name=windows-exporter
Annotations: <none>
Subsets:
Addresses: 172.31.249.177,172.31.249.147
NotReadyAddresses: <none>
Ports:
Name Port Protocol
---- ---- --------
metrics 9182 TCP
Events: <none>
"version": "3.1.0+0a3a937"
Bug tested mistakenly on 4.9 + 3.1.0 Testing on OCP 4.9 WMCO 4.0.0+2f0b49a2 he endpoint 10.0.128.191 has been deleted, yet still exist oc describe endpoints -n openshift-windows-machine-config-operator Name: windows-exporter Namespace: openshift-windows-machine-config-operator Labels: name=windows-exporter Annotations: <none> Subsets: Addresses: 10.0.128.191,10.0.140.91,10.0.143.88 NotReadyAddresses: <none> @rrasouli Can you share the WMCO logs associated with the failed test? 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 (Windows Container Support for Red Hat OpenShift 4.0.0 product release), 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-2021:3702 |
Description of problem: In vSphere, WMCO miss the `Deleting` phase notification event, leaving incorrect node's information in the `windows-exporter` metrics endpoint, where `Subsets` contains the IP address of a machine that is no longer available, resulting in an invalid mapping for Prometheus metrics endpoint. It’s important to note, that this bug applies only for vSphere platform, given how quickly virtual machines are destroyed in the vCenter. In AWS, for example, the EC2 instances spend more time in `Deleting` phase allowing WMCO to fully process all the events. Version-Release number of selected component (if applicable): WMCO 2.0.3 running on cluster with version 4.7.23 WMCO 3.0.0 running on cluster with version 4.8.4 How reproducible: Sometimes, depends on platform performance while removing a virtual machine Steps to Reproduce: 1. WMCO configured and running 2. Create a valid machineSet with 1 replicas 3. Observe the node information in the `windows-exporter` metrics endpoint object. Note the IP Addresses, for example: 172.31.251.250 4. Delete the machineSet 5. Wait for the Windows machine to disappear 6. Check one more time the `windows-exporter` metrics endpoint object, if there is still an entry in `Subsets` mapping an IP Address of a deleted machine, you have reproduced the bug. Metrics are no longer available for a deleted machine Bug frequency is about 0.75 (6 of 8 runs), with replicas set to 1. In step #2, you can set replicas to 5 to increase the bug frequency rate. Actual results: The `windows-exporter` metrics endpoint object contains Subsets with an IP Address of a deleted machine ``` $ oc describe endpoints -n openshift-windows-machine-config-operator Name: windows-exporter Namespace: openshift-windows-machine-config-operator Labels: name=windows-exporter Annotations: <none> Subsets: Addresses: 172.31.251.250 NotReadyAddresses: <none> Ports: Name Port Protocol ---- ---- -------- metrics 9182 TCP Events: <none> ``` Expected results: The IP Address of the deleted machine does not appears in the `windows-exporter` metrics endpoint object. With replicas set to 1, the Subsets must have no entries, empty. ``` $ oc describe endpoints -n openshift-windows-machine-config-operator Name: windows-exporter Namespace: openshift-windows-machine-config-operator Labels: name=windows-exporter Annotations: <none> Subsets: Events: <none> ```