Fluentd Kinesis Firehose Helm Chart

Fluentd Kinesis Firehose Helm Chart creates a Kubernetes DaemonSet and stream the logs to Amazon Kinesis Firehose

Getting started

Requirements

The Fluentd kinesis Firehose daemonset requires that an AWS account has already been provisioned with a Kinesis Firehose stream and with its data stores (eg. Amazon S3 bucket, Amazon Elasticsearch Service, etc). Available is a CX Cloud provided Terraform module, terraform-kinesis-firehose-elasticsearch for helping with the installation.

Additionally there should already be a running Kubernetes cluster.

Installing the chart

Clone the repository:

git clone [email protected]:cxcloud/helm-fluentd-kinesis-firehose.git

To install the chart with the release name my-release into the namespace kube-system:

helm install helm-fluentd-kinesis-firehose --name my-release --namespace kube-system

Uninstalling the Chart

To uninstall/delete the my-release deployment:

helm delete --purge my-release

Configuration

The following table list the configurable parameters of the chart and their default values.

Parameter	Description	Default
image.repository	Repository for docker image	fluent/fluentd-kubernetes-daemonset
image.tag	Docker image tag	v0.12-debian-kinesis
image.pullPolicy	Image pull policy	IfNotPresent
nameOverride	Override the name for resources	Empty string
fullnameOverride	Override the full name for resources	Empty string
fluentEnvs.assumeRole	Should Fluentd assume IAM role for accessing Kinesis	false
fluentEnvs.roleARN	AWS IAM role	Empty string
fluentEnvs.roleSession	Role session	Empty string
fluentEnvs.kinesisRegion	AWS region for Kinesis	eu-west-1
fluentEnvs.kinesisStreamName	Kinesis stream name	cxcloud
fluentEnvs.kinesisIncludeTimeKey	Include time key	true
fluentEnvs.bufferChunkLimitSize	Buffer chunk limit size	2M
fluentEnvs.bufferQueueLimitLength	Buffer queue limit length	32
fluentConf	fluent.conf content	Not set
systemdConf	systemd.conf content	Not set
kubernetesConf	kubernetes.conf content	Not set
resources	Resources allocation (Requests and Limits)	{}
rbac.create	Whether RBAC resources are created	true
instanceProfileCredentials	AWS instance profile credentials	false
instanceProfileIPAddress	IP Address to get instance profile credentials	169.254.169.254
instanceProfilePort	Port to get instance profile credentials	80
podAnnotations	Annotations	{}

Overriding configurations

Parameters can be overridden when installing the chart. For example assuming an IAM role for another AWS account were Kinesis Firehose is running:

helm install helm-fluentd-kinesis-firehose --name my-release --namespace kube-system \
  --set fluentEnvs.useRole=true \
  --set fluentEnvs.roleARN=arn:aws:iam::012345678901:role/KinesisFirehose \
  --set fluentEnvs.roleSession=kops-nodes

Alternatively, a YAML file (values.yaml) that specifies the values for the parameters can be provided while installing the chart:

helm install helm-fluentd-kinesis-firehose --name my-release --namespace kube-system -f values.yaml

Fluentd DaemonSet

The daemonset is using the fluentd-kubernetes-daemonset Docker image for Kinesis as the base.

fluent.conf

The Kinesis Docker image contains preset configuration files for Kinesis Data stream that is not compatible with Kinesis Firehose. However, the image is using the Fluent plugin for Amazon Kinesis with support for all Kinesis services. Hence, fluent.conf has to be overwritten by a custom configuration file in order to work with Kinesis Firehose.

kubernetes.conf

The Docker image itself contains a pre-configured configuration, kubernetes.conf. The content can be overwritten from the kubernetesConf variable in configurations.

systemd.conf

The Docker image itself contains a pre-configured configuration, systemd.conf. The content can be overwritten from the systemdConf variable in configurations.

parse-json.conf

Kubernetes log stdout from all the containers into log files. These log files contain JSON strings containing the container log line in a "log" attribute. In case the container log is already in JSON formant the parse-json.conf will parse the json object into own key/value pairs. This is useful for example if Elasticsearch is used as a data source for Kinesis Firehose.

License

MIT © CXCloud