bhouse-nexthop opened a new issue, #14628:
URL: https://github.com/apache/cloudstack/issues/14628
### problem
A CKS cluster created before 4.21 can't be scaled up after CloudStack is
upgraded to 4.21 or later. `scaleKubernetesCluster` with a larger `size` fails
within seconds, with no VM created:
```
Unexpected exception while executing
org.apache.cloudstack.api.command.user.kubernetes.cluster.ScaleKubernetesClusterCmd
java.lang.NullPointerException: Cannot invoke
"com.cloud.template.VirtualMachineTemplate.getUniqueName()" because "template"
is null
at
com.cloud.vm.VirtualMachineManagerImpl.isBlankInstanceDefaultTemplate(VirtualMachineManagerImpl.java)
at
com.cloud.vm.VirtualMachineManagerImpl.isBlankInstance(VirtualMachineManagerImpl.java)
at
com.cloud.vm.UserVmManagerImpl.createAdvancedVirtualMachine(UserVmManagerImpl.java)
at
com.cloud.kubernetes.cluster.actionworkers.KubernetesClusterResourceModifierActionWorker.createKubernetesNode(KubernetesClusterResourceModifierActionWorker.java)
at
com.cloud.kubernetes.cluster.actionworkers.KubernetesClusterResourceModifierActionWorker.provisionKubernetesClusterNodeVms(KubernetesClusterResourceModifierActionWorker.java)
at
com.cloud.kubernetes.cluster.actionworkers.KubernetesClusterScaleWorker.scaleUpKubernetesClusterSize(KubernetesClusterScaleWorker.java)
```
The cluster goes back to Running.
Cause:
- #9102 (4.21) added `control_node_template_id`, `worker_node_template_id`
and `etcd_node_template_id` to `kubernetes_cluster`. The upgrade adds the
columns but doesn't fill them in, so they are NULL for every cluster created
before 4.21.
- `KubernetesClusterActionWorker.init()` looks the node templates up from
those columns only
(`templateDao.findById(kubernetesCluster.getWorkerNodeTemplateId())`), which
gives null.
- `createKubernetesNode` passes that null template to
`createAdvancedVirtualMachine`.
- Before 4.21, nodes were deployed from the template CKS picks by default
(`clusterTemplate`), and the owner was given launch permission for it during
the operation. `isDefaultTemplateUsed()`, which decides about that launch
permission, also returns false for these clusters, since none of the three
columns equals `template_id`.
### versions
4.21.0.0 and later, including 4.22 and main, for clusters created on 4.20 or
earlier.
### The steps to reproduce the bug
1. On 4.20, create a CKS cluster and let it start.
2. Upgrade CloudStack to 4.22.
3. Scale the cluster up by one worker (`scaleKubernetesCluster id=<cluster>
size=<current + 1>`).
4. The job fails with the NullPointerException above. The cluster's
`control_node_template_id`, `worker_node_template_id` and
`etcd_node_template_id` are NULL.
Setting `control_node_template_id` and `worker_node_template_id` to the
cluster's `template_id` in the database works around it.
### What to do about it?
When a node type has no template set, deploy its nodes from the default CKS
template, as before 4.21, and count the cluster as using the default template
so the owner gets launch permission for it. Filling the columns in during the
upgrade would be a schema change; the code fallback also covers clusters that
were already upgraded.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
