Does ClusterLogging have a plan to support fluentd input and output configurations on a large scale?
Closed this issue · 2 comments
I'm using cluster-logging 5.4.3, and I want to configure fluentd's configuration file independently, but when I look at the ClusterLogging and ClusterLogForwarder fields, fluentd doesn't provide many configuration fields. It is recommended that managementState be changed to Unmanaged. In this way, fluentd.conf can be set. Does ClusterLogging have a plan to support fluentd input and output configurations on a large scale?
When I use such a YAML file, the fluentd.conf file does not collect the corresponding collection path.Look at the pod status.
apiVersion: "logging.openshift.io/v1"
kind: "ClusterLogging"
metadata:
name: "instance"
namespace: "openshift-logging"
spec:
managementState: "Managed"
collection:
logs:
type: "fluentd"
fluentd:
resources:
limits:
memory: 736Mi
requests:
cpu: 100m
memory: 736Mi
forwarder:
fluentd:
inFile:
readLinesLimit: 1000
path: /home/fang/test/*.log
tag: fluentd.test
Pod state:
apiVersion: logging.openshift.io/v1
kind: ClusterLogging
metadata:
creationTimestamp: "2023-04-28T03:15:35Z"
generation: 1
name: instance
namespace: openshift-logging
resourceVersion: "491193908"
uid: 332f7da3-cb70-4939-9d06-929acd6a44ad
spec:
collection:
logs:
fluentd:
resources:
limits:
memory: 736Mi
requests:
cpu: 100m
memory: 736Mi
type: fluentd
forwarder:
fluentd:
inFile:
readLinesLimit: 1000
managementState: Managed
status:
clusterConditions:
- lastTransitionTime: "2023-04-28T03:15:55Z"
message: ClusterLogForwarder input validation error
reason: ClusterLogForwarder input validation error
status: "True"
I want to konw how can I configure fluentd's configuration file independently.
The intent of this solution is to provide an abstraction to log collection and storage and limiting the configuration details that are surfaced in the API. It is not another layer over fluentd configuration to make it "kubernetes" like. If you have requirements to control larger aspects of the configuration then I encourage you to create and manage your own deployment outside of the operator.
We have deprecated fluentd in favor of vector and in doing so have made it possible to swap the collector with almost a 1 for 1 replacement. Given we are prioritizing fluentd, we have effectively stopped any further development.