Elastic-Agent autodiscovery for Kubernetes pod logs not working after upgrading to 8.6.x

Elastic Stack Elastic Agent
1

Hi all,

I noticed an issue with the standalone Elastic Agent shipping Kubernetes pod logs after upgrading the Agent version from 8.5.3 to 8.6.x.

We run the Elastic-Agent standalone on Kubernetes as a DaemonSet.

After deploying the version 8.6.x (tested with 8.6.0 and 8.6.1) logs for some pods are not being shipped to elasticsearch. When I restart the affected elastic-agents they start working correctly.

However when a application pod is redeployed the logs of the new pod are not shipped as well, until the agent is restarted manually.

I tested with both types filestream and logfile.

This is the agent policy inputs config we are using for the kubernetes pod logs:

  - id: kubernetes-application-container-logs
    name: kubernetes-application-container-logs
    revision: 1
    #type: filestream
    type: logfile
    use_output: default
    meta:
      package:
        name: kubernetes
        version: 1.31.2
    data_stream:
      namespace: applications
    streams:
      - id: kubernetes-application-logs-${kubernetes.pod.name}-${kubernetes.container.id}
        data_stream:
          dataset: kubernetes.container
          type: logs
        paths:
          - '/var/log/containers/*${kubernetes.container.id}.log'
        #prospector.scanner.symlinks: true
        symlinks: true
        pipeline: logs-kubernetes-pipeline
        condition: ${kubernetes.namespace} != 'kube-system'
        # parsers:
        #   - container:
        #       stream: all
        #       format: auto
        processors:
          - add_fields:
                target: ''
                fields:
                  environment: abc
                  cluster: xyz

As the same config works fine on Elastic-Agent 8.5.3, I assume there's a bug in the new version.

Steps to reproduce:

  1. Deploy Elastic-Agent version 8.6.0 as standalone on Kubernetes as daemonset.
  2. Check if Kubernetes application pod logs are shipped to Elasticsearch.
  3. Delete a kubernetes application pod.
  4. Wait for Kubernetes to rededeploy the pod.
  5. Check that the logs from the new pod are not shipped to Elasticsearch.

Can anybody confirm this behavior/bug? Do you have a solution for that?

Thanks!

2

For anybody reading this post having the same issue, it's going to be tracked here: Elastic Agent 8.6.x standalone deployment in Kubernetes doesn't start monitoring new pods until agent restart · Issue #2269 · elastic/elastic-agent · GitHub

3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.