Created attachment 1271571 [details] ansible inventory Description of problem: I installed a OCP 3.5 cluster via ansible advanced installer and configured it to roll out metrics + logging with it, using NFS shares as storage. Once the rollout has taken place, the PVs for metrics and logging are swapped. Version-Release number of selected component (if applicable): OpenShift Master: v3.5.5.5 Kubernetes Master: v1.5.2+43a9be4 How reproducible: 50% - it seems to be random whether metrics and logging PVs are correct or swapped Steps to Reproduce: 1. configure ansible inventory like in the attached sample 2. run installer 3. check assignment with oc get pv Actual results: ovpn-116-18:~ wolfram$ oc get pv NAME CAPACITY ACCESSMODES RECLAIMPOLICY STATUS CLAIM REASON AGE logging-volume 10Gi RWO Retain Bound openshift-infra/metrics-1 12m metrics-volume 10Gi RWO Retain Bound logging/logging-es-0 12m Expected results: logging-volume is claimed in the logging project and metrics-volume is claimed in the openshift-infra project Additional info:
marking this as low severity. Things work. NFS isn't supported for metrics as I understand it. Now if we do want to fix it, we could/should partially fill in the claimRef when creating the PV. That way the binder will always find the right PVC. In any case, other than it looks silly, things should be working just fine...
(In reply to Eric Paris from comment #1) > marking this as low severity. Things work. NFS isn't supported for metrics > as I understand it. According to [1] and [2], seem like our offical doc does not claim that NFS is not supported for logging and metrics. If we really not support that, we should remove all NFS storage from the above doc. [1]: https://github.com/openshift/openshift-ansible/blob/master/inventory/byo/hosts.ose.example#L472 [2]: https://docs.openshift.com/container-platform/3.5/install_config/install/advanced_install.html#metrics-storage
# oc get pv NAME CAPACITY ACCESSMODES RECLAIMPOLICY STATUS CLAIM STORAGECLASS REASON AGE logging-volume 10Gi RWO Retain Bound logging/logging-es-0 44m metrics-volume 10Gi RWO Retain Bound openshift-infra/metrics-1 44m But I tested on # openshift version openshift v3.6.74 kubernetes v1.6.1+5115d708d7 etcd 3.1.0 It is not the latest OCP 3.6.0 puddle, will re-test tomorrow, but I believe this issue is fixed.
Tested again with latest OCP 3.6.0 puddle, issue fixed,it did not confuse PVs for metrics and logging # oc get pv NAME CAPACITY ACCESSMODES RECLAIMPOLICY STATUS CLAIM STORAGECLASS REASON AGE logging-volume 10Gi RWO Retain Bound logging/logging-es-0 55m metrics-volume 10Gi RWO Retain Bound openshift-infra/metrics-1 55m # openshift version openshift v3.6.121 kubernetes v1.6.1+5115d708d7 etcd 3.2.0 Set it to VERIFIED
logging and metrics images from ops mirror # docker images | grep logging logging-elasticsearch v3.6 41d18242fe2e 2 hours ago 404.5 MB logging-kibana v3.6 1f9fde5cd870 2 hours ago 342.4 MB logging-fluentd v3.6 7e35b19b4e40 2 hours ago 232.5 MB logging-curator v3.6 a0148dd96b8d 12 days ago 221.5 MB # docker images | grep metrics metrics-hawkular-metrics v3.6 fe12100d5533 About an hour ago 1.293 GB metrics-cassandra v3.6 500536a9d23e 2 hours ago 572.8 MB metrics-heapster v3.6 821075975cc9 2 hours ago 274.4 MB
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/RHEA-2017:1716