@steffens I've just upgraded to version 6.4.0, previously I was using version 6.3.2 and I didn't have this issue. On the other hand in 6.3.2 I had issue where logs longer then 16kB were split up (due to docker's json-file limits). 6.4.0 Fixed this issue and it looks like CRI issue originates from the same code block.
I don't think it's multiple prospectors reading the same file.
Here's the config:
filebeat.autodiscover:
providers:
- type: kubernetes
in_cluster: true
hints.enabled: true
include_annotations: '*'
processors:
- add_cloud_metadata: ~
- add_kubernetes_metadata:
in_cluster: true
- add_docker_metadata:
output.elasticsearch:
hosts: ["http://elasticsearch:9200"]
http.enabled: true
http.port: 5066