CODEWITHSUNDEEP
kubernetes
Đinh Anh Huy
Đinh Anh Huy

Reputation: 545

How to kubectl get event of node (not pod)?

When describing a node, there are history conditions that show up.

Conditions:
  Type                 Status  LastHeartbeatTime                 LastTransitionTime                Reason                       Message
  ----                 ------  -----------------                 ------------------                ------                       -------
  NetworkUnavailable   False   Tue, 10 Aug 2021 10:55:23 +0700   Tue, 10 Aug 2021 10:55:23 +0700   CalicoIsUp                   Calico is running on this node
  MemoryPressure       False   Mon, 16 Aug 2021 12:02:18 +0700   Thu, 12 Aug 2021 14:55:48 +0700   KubeletHasSufficientMemory   kubelet has sufficient memory available
  DiskPressure         False   Mon, 16 Aug 2021 12:02:18 +0700   Thu, 12 Aug 2021 14:55:48 +0700   KubeletHasNoDiskPressure     kubelet has no disk pressure
  PIDPressure          False   Mon, 16 Aug 2021 12:02:18 +0700   Thu, 12 Aug 2021 14:55:48 +0700   KubeletHasSufficientPID      kubelet has sufficient PID available
  Ready                False   Mon, 16 Aug 2021 12:02:18 +0700   Mon, 16 Aug 2021 11:54:02 +0700   KubeletNotReady              PLEG is not healthy: pleg was last seen active 11m17.462332922s ago; threshold is 3m0s

I have 2 questions:

  1. I think those conditions only show the latest status. How can I access the full-time history of the previous conditions?
  2. Suggest me the tool that converts node condition to something like pod events for log centralize.

Upvotes: 6

Views: 7603

Answers (3)

neoakris
neoakris

Reputation: 5145

kubectl get events --field-selector involvedObject.kind=Node,type!=Normal --sort-by='.lastTimestamp'

Note: you can combine multiple field-selectors by using commas. I came up with this gem that's super useful for debugging clusters with 100's of nodes, as it only shows relevant data !=Normal filters the kruft and sorts it by time.

Upvotes: 0

matt_j
matt_j

Reputation: 4614

You're right, the kubectl describe <NODE_NAME> command shows the current condition status (False/True).

You can monitor Nodes events using the following command:

# kubectl get events --watch --field-selector involvedObject.kind=Node 
LAST SEEN   TYPE      REASON                 OBJECT         MESSAGE
3m50s       Warning   EvictionThresholdMet   node/kworker   Attempting to reclaim inodes
44m         Normal    NodeHasDiskPressure    node/kworker   Node kworker status is now: NodeHasDiskPressure

To view only status related events, you can use grep with the previous command:

# kubectl get events --watch --field-selector involvedObject.kind=Node | grep "status is now"
44m         Normal    NodeHasDiskPressure    node/kworker   Node kworker status is now: NodeHasDiskPressure

By default, these events are retained for 1 hour. However, you can run the kubectl get events --watch --field-selector involvedObject.kind=Node command from within a Pod and collect the output from that command using a log aggregation system like Loki. I've described this approach with a detailed explanation here.

Upvotes: 9

Chandra Sekar
Chandra Sekar

Reputation: 763

You can use kubectl describe node <nodename> --show-events=trueand kubectl get events which will show you the events related to the described object which are persisted in etcd and provide high-level information on what is happening in the cluster.

Upvotes: 1

Related Questions