Hide Forgot
Description of problem: After changing master-config.yaml to enlarge the clusternetwork CIDR, admin cannot create the hostsubnet for f5. Version-Release number of selected component (if applicable): openshift v3.4.0.15+9c963ec kubernetes v1.4.0+776c994 etcd 3.1.0-alpha.1 How reproducible: Always Steps to Reproduce: 1. make sure the hostsubnet for f5 can be created with default setting, then delete it. 2. vi /etc/origin/master/master-config.yaml and change the clusternetwork CIDR to: clusterNetworkCIDR: 10.0.0.0/14 ## default is 10.1.0.0/16 3. restart master and check the clusternetwork [root@ip-172-18-12-194 ~]# oc get clusternetwork NAME NETWORK HOST SUBNET LENGTH SERVICE NETWORK PLUGIN NAME default 10.0.0.0/14 8 172.30.0.0/16 redhat/openshift-ovs-multitenant [root@ip-172-18-12-194 ~]# 4. Create the hostsubnet again. oc create -f f5host.yaml (see additiontal info for file content) Actual results: It shows hostsubnet "f5-server" created but cannot find it when getting the resource. [root@ip-172-18-12-194 ~]# oc create -f f5host.yaml hostsubnet "f5-server" created [root@ip-172-18-12-194 ~]# [root@ip-172-18-12-194 ~]# oc get hostsubnet NAME HOST HOST IP SUBNET ip-172-18-12-194.ec2.internal ip-172-18-12-194.ec2.internal 172.18.12.194 10.1.2.0/24 ip-172-18-9-101.ec2.internal ip-172-18-9-101.ec2.internal 172.18.9.101 10.0.0.0/24 ip-172-18-9-102.ec2.internal ip-172-18-9-102.ec2.internal 172.18.9.102 10.1.0.0/24 [root@ip-172-18-12-194 ~]# [root@ip-172-18-12-194 ~]# Expected results: The hostsubnet for f5 can be created after changing the clusternetwork CIDR. Additional info: <f5host.yaml> apiVersion: v1 host: f5-server hostIP: 10.3.89.213 kind: HostSubnet metadata: annotations: pod.network.openshift.io/assign-subnet: "true" name: f5-server
If this needs manual steps, there's a docs PR on how to resize things open: https://github.com/openshift/openshift-docs/pull/3112
Could not reproduce the problem. The hostsubnet gets created just fine and can be viewed too. Do you have any logs from the master when this is observed?
not sure the root cause, but tried the same steps also cannot reproduce it in OCP 3.4.0.18.
There have been a bunch of fixes to networking recently, so I'm going to mark this as unreproducible. Please re-open if you see it again.