This message was deleted.
# logginga
adamant-kite-43734
10/19/2024, 4:44 PMThis message was deleted.
p
powerful-librarian-10572
10/21/2024, 7:27 AMThis is only relevant for the logs managaed by kubelets for use by kubectl and such. If you build an ELK stack, then the data will be saved there as well and not rotated.
l
late-air-9846
10/21/2024, 11:56 AMif the logs cannot be processed by the log collector (fluentbit) soon enough, then the logs are getting lost indeed
a
acceptable-jackal-29556
10/22/2024, 5:16 AMWhat can we do to mitigate the same... and are logs of evicted pods removed from file system
acceptable-jackal-29556
10/22/2024, 5:17 AM@powerful-librarian-10572 is there a possibility that the logs of an evicted pod are deleted before the fluent bit collector runs. I agree that once fluentbit collects and sends to ELK it is safe
p
powerful-librarian-10572
10/22/2024, 7:47 AMI mean 0 risk doesnt exist lmao.
powerful-librarian-10572
10/22/2024, 7:48 AMHowever, pod eviction should only occur on node event, therefore those logs might not be of high importance anyway. If you are scared about pod logs on crash not being collected, this shouldnt occur.
l
late-air-9846
10/22/2024, 7:53 AMeviction can happen also if the scheduler needs resources for a higher priority pod, but if you have workloads where logs are critical, then use guaranteed QoS or priorityclasses, etc
p
powerful-librarian-10572
10/22/2024, 7:54 AMYeah thats what i mean by node event, a pod wont get evicted by its own request 😄
l
late-air-9846
10/22/2024, 7:54 AMright
p
powerful-librarian-10572
10/22/2024, 7:55 AMl
late-air-9846
10/22/2024, 8:16 AMloki is not easy to operate
🫠1