-
Notifications
You must be signed in to change notification settings - Fork 663
Description
I have an issue with setting a node in "NotReady", when the FileSystemIsReadOnly.
I tried to illustrate the scenario to test readOnly condition on node.
steps:
NPD: v0.8.20
- configure node-problem-detector: with extending --config.system-log-monitor=.../config/readonly-monitor.json
on description I see that the node condtions is set as below:
Conditions:
Type Status Reason Message
...
ReadonlyFilesystem False FilesystemIsNotReadOnly Filesystem is not read-only
....
Ready True KubeletReady kubelet is posting ready status
- on specific node I tried to send an information to /dev/kmsg that "Remounting filesystem read-only",which caused that ReadonlyFilesystem will be true, and it was changed to True
On description this node I got those info:
Conditions:
Type Status LastHeartbeatTime LastTransitionTime Reason Message
ReadonlyFilesystem True FilesystemIsReadOnly Remounting filesystem read-only
Ready True KubeletReady kubelet is posting ready status
Warning FilesystemIsReadOnly 24h readonly-monitor Node condition ReadonlyFilesystem is now: True, reason: FilesystemIsReadOnly, message: "Remounting filesystem read-only"
But Node still in status Ready.
when I tried to get
curl localhost:8080/api/v1/nodes/nodeName/status
I got in conditions section:
{
"type": "ReadonlyFilesystem",
"status": "True",
"reason": "FilesystemIsReadOnly",
"message": "Remounting filesystem read-only"
},
Question is: Why Node still Ready, and how does kube-apiserver report this info, when NDP is setting this flag as True on NodeCondition?
Do I miss something in configuration to enforce that node should be NotReady?
and how does the best practice config look like?