事前準備

S3をバックエンドにしたFirehoseリソースを作成しておく

  FirehoseApplicationLogRole:
    Type: AWS::IAM::Role
    Properties:
      RoleName: my-sweet-firehose-application-log-role
      AssumeRolePolicyDocument:
        Version: '2012-10-17'
        Statement:
          - Action: sts:AssumeRole
            Effect: Allow
            Principal:
              Service: firehose.amazonaws.com
      Policies:
        - PolicyDocument:
            Version: '2012-10-17'
            Statement:
              - Action:
                  - glue:GetTable
                  - glue:GetTableVersion
                  - glue:GetTableVersions
                Effect: Allow
                Resource: '*'
              - Action:
                  - s3:AbortMultipartUpload
                  - s3:GetBucketLocation
                  - s3:GetObject
                  - s3:ListBucket
                  - s3:ListBucketMultipartUploads
                  - s3:PutObject
                Effect: Allow
                Resource:
                  - my-sweet-s3-bucket
                  - my-sweet-s3-bucket/*
          PolicyName: my-sweet-firehose-application-log-policy

  FirehoseApplicationLog:
    Type: AWS::KinesisFirehose::DeliveryStream
    Properties:
      DeliveryStreamName: "my-sweet-deliver-stream" # max 64 chars
      DeliveryStreamType: "DirectPut"
      S3DestinationConfiguration:
        BucketARN: "my-sweet-s3-bucket"
        BufferingHints:
          IntervalInSeconds: 60 # Default: 300
        CompressionFormat: 'GZIP'
        EncryptionConfiguration:
          NoEncryptionConfig: NoEncryption # Encryption is done by S3
        RoleARN: !Ref FirehoseApplicationLogRole

EKSのログをFluentdでFirehoseに送る

次の fluentd-firehose.yaml を k8s にデプロイする。CloudWatch Container Insights の fluentd config を参考に書いた。

Firehoseでログ収集したい deployment に fluentd_firehose_delivery_stream_name という annotation で Firehose の Delivery Stream 名を指定すると転送される。

apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-sweet-api
spec:
  template:
    metadata:
      annotations:
        fluentd_firehose_delivery_stream_name: "my-sweet-firehose-stream-name"

Athenaテーブルを作成する

CREATE DATABASE application_log

CREATE EXTERNAL TABLE application_log.my_sweet_api (log string, stream string, docker string, kubernetes string) PARTITIONED BY (date string) ROW FORMAT SERDE 'org.openx.data.jsonserde.JsonSerDe' LOCATION 's3://my-sweet-s3-bucket' TBLPROPERTIES ( 'projection.enabled' = 'true', 'projection.date.type' = 'date', 'projection.date.range' = '2020/01/01,NOW', 'projection.date.format' = 'yyyy/MM/dd', 'projection.date.interval' = '1', 'projection.date.interval.unit' = 'DAYS', 'storage.location.template' = 's3://my-sweet-s3-bucket/${!date}')

Athenaにクエリを投げる

log フィールドにコンテナログが入る。JSONなログであれば json_extract_scalar を利用して取り出せる。
Partitioned table にしてあるのでパーティションキーである date で絞り込むと安くて速い。

SELECT
  json_extract_scalar(log, '$.severity') AS severity,
  json_extract_scalar(log, '$.timestamp') AS timestamp,
  json_extract_scalar(kubernetes, '$.container_name') AS container_name,
  json_extract_scalar(kubernetes, '$.namespace_name') AS namespace_name
FROM application_log.my_sweet_api
WHERE
  date = '2021/01/21'
  AND from_iso8601_timestamp(json_extract_scalar(log, '$.timestamp')) BETWEEN timestamp '2021-01-21 18:42:00 Asia/Tokyo' AND timestamp '2021-01-21 18:44:00 Asia/Tokyo'
LIMIT 50;