Red Hat Bugzilla – Bug 1339154
Dynamic provisioned pv/volume are not deleted after pod/pvc are deleted.
Last modified: 2017-03-08 13:26 EST
Description of problem: GCE dynamic provisioned pv/volume are not deleted after pod/pvc are deleted. PV changed to Released status, but are not deleted. And volume remains attached to nodes. Version-Release number of selected component (if applicable): openshift v3.2.0.44 kubernetes v1.2.0-36-g4a3f9c5 etcd 2.2.5 How reproducible: Always Steps to Reproduce: 1.Create dynamic PVC. 2.Check PV/PVC are bound. 3.Create pod that use the PVC. 4.Delete the pod and PVC. 5.Check PV and volumes. Actual results: PV changed to released status, but are not deleted. Volumes remains attached to nodes. # oc get pv NAME CAPACITY ACCESSMODES STATUS CLAIM REASON AGE pv-gce-4r4h8 3Gi ROX Released wu2hl/gcec3-wu2hl 25m pv-gce-63jz7 3Gi ROX Released uyxke/gcec3-uyxke 17m pv-gce-75r33 2Gi RWX Released uyxke/gcec2-uyxke 17m pv-gce-cs9zb 1Gi RWO Released uyxke/gcec1-uyxke 17m pv-gce-nw2wk 1Gi RWO Released wu2hl/gcec1-wu2hl 25m pv-gce-rkxd2 2Gi RWX Released wu2hl/gcec2-wu2hl 25m regpv-volume 17G RWX Bound default/regpv-claim 6h # oc describe pv pv-gce-4r4h8 Name: pv-gce-4r4h8 Labels: Status: Released Claim: wu2hl/gcec3-wu2hl Reclaim Policy: Delete Access Modes: ROX Capacity: 3Gi Message: Source: Type: GCEPersistentDisk (a Persistent Disk resource in Google Compute Engine) PDName: kubernetes-dynamic-pv-gce-4r4h8 FSType: ext4 Partition: 0 ReadOnly: false Expected results: PV/volumes are deleted after pod/PVC are deleted. Additional info:
on node qe-lxia-2401-node-container-registry-router-2 # mount | grep dynamic /dev/sdb on /var/lib/origin/openshift.local.volumes/plugins/kubernetes.io/gce-pd/mounts/kubernetes-dynamic-pv-gce-nw2wk type ext4 (rw,relatime,seclabel,data=ordered) /dev/sdc on /var/lib/origin/openshift.local.volumes/plugins/kubernetes.io/gce-pd/mounts/kubernetes-dynamic-pv-gce-4r4h8 type ext4 (rw,relatime,seclabel,data=ordered) /dev/sdd on /var/lib/origin/openshift.local.volumes/plugins/kubernetes.io/gce-pd/mounts/kubernetes-dynamic-pv-gce-rkxd2 type ext4 (rw,relatime,seclabel,data=ordered) /dev/sde on /var/lib/origin/openshift.local.volumes/plugins/kubernetes.io/gce-pd/mounts/kubernetes-dynamic-pv-gce-cs9zb type ext4 (rw,relatime,seclabel,data=ordered) /dev/sdf on /var/lib/origin/openshift.local.volumes/plugins/kubernetes.io/gce-pd/mounts/kubernetes-dynamic-pv-gce-75r33 type ext4 (rw,relatime,seclabel,data=ordered)
I can reproduce this too on my containerized environment Steps to Reproduce: 1.Create dynamic PVC. 2.Check PV/PVC are bound. 3.Create pod that use the PVC. 4.Delete the pod and PVC. 5.Check PV and volumes. When the pod was deleted in step 4, I could see the volume was still mounted to my node(running `mount|grep gce`), the PV could not be deleted because the volume was still attached to the instance.
Happens to aws/cinder too
Indeed, Kubernetes cannot delete a PV that is still attached somewhere. Attaching was completely rewritten in Kubernetes 1.3 (waiting to be rebased in OpenShift 3.3) and this bug should be already gone there. Do you have a script that can reproduce it? I could run it with current Kubernetes to be sure. Liang, Hou, does it happen *only* to containerized OpenShift or also one installed from RPM? Also, does this happen always or does it break only sometimes?
Always happen on containerized OpenShift. Fine on RPM environment.
This seems to be fixed in Kubernetes 1.3. It is *not* fixed in OpenShift Origin v1.3.0-alpha.2 (it was rebased to an older Kubernetes 1.3-alpha).
Verified on below version, openshift v3.3.0.16 kubernetes v1.3.0+507d3a7 etcd 2.3.0+git The PV and volume are deleted in a short time(less than 10 sec) after delete pod/pvc.
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-2016:1933