oc adm drain --ignore-daemonsets seems to not ignore daemonsets

[dustymabe]

oc adm drain --ignore-daemonsets seems to not ignore daemonsets

Version

[root@origin-master-1 ~]# oc version
oc v3.6.1+008f2d5
kubernetes v1.6.1+5115d708d7
features: Basic-Auth GSSAPI Kerberos SPNEGO

Server https://y.y.y.y:8443
kubernetes v1.6.1+5115d708d7

steps:

[root@origin-master-1 ~]# oc adm drain x.x.x.x
node "x.x.x.x" already cordoned
error: pods with local storage (use --delete-local-data to override): glusterfs-storage-5h928; DaemonSet-managed pods (use --ignore-daemonsets to ignore): glusterfs-storage-5h928
[root@origin-master-1 ~]# oc adm drain --ignore-daemonsets x.x.x.x
node "x.x.x.x" already cordoned
error: pods with local storage (use --delete-local-data to override): glusterfs-storage-5h928
[root@origin-master-1 ~]# oc get ds --namespace glusterfs
NAME                DESIRED   CURRENT   READY     UP-TO-DATE   AVAILABLE   NODE-SELECTOR            AGE
glusterfs-storage   3         3         3         3            3           glusterfs=storage-host   26d
$ oc get pods -o wide
NAME                      READY     STATUS    RESTARTS   AGE       IP                NODE
glusterfs-storage-5h928   1/1       Running   2          26d       x.x.x.x            x.x.x.x
glusterfs-storage-vd02b   1/1       Running   2          26d       x.x.x.y            x.x.x.y
glusterfs-storage-z7q5r   1/1       Running   3          26d       x.x.x.z            x.x.x.z
heketi-storage-1-wpfb3    1/1       Running   0          9d        x.x.x.y            x.x.x.y

[dustymabe]

note I was able to workaround by using a complicated pod-selector:

[root@origin-master-1 ~]# oc adm manage-node --evacuate --force x.x.x.x --pod-selector='glusterfs notin (storage-pod)'                                                                                                                
Flag --evacuate has been deprecated, use 'oadm drain NODE' instead

Migrating these pods on node:
<snip>
....

[juanvallejo]

Hi Dusty, are you by any chance able to reproduce this with a 3.7 or 3.8 client?

[juanvallejo]

An immediate workaround for draining a node with pods managed by a DaemonSet that also have local storage, is to use both the --ignore-daemonsets and --delete-local-data flags. Having both flags causes the command to succeed, and will in fact leave any daemonset-managed pods (and their storage) intact, as those are never deleted by the drain command.

[juanvallejo]

Related issue: #17563