Reading from object storage

Within the kdb Insights Database, the Data Access service can be configured to read directly from object storage. This mode allows a kdb+ partitioned database to be read from a static location in object storage. This is useful for easily querying existing kdb data that is managed outside of kdb Insights.

Writing to object storage

Reading directly from object storage uses read-only data. To write data to object storage, use an object storage tier.

Tutorial

Query a kdb+ partitioned database from object storage

This example will deploy a kdb Insights Database that reads static data from an AWS S3 bucket. The same process works for Azure blob storage and Google Cloud Storage. Define the appropriate authentication parameters based on the selected cloud vendor.

Uploading your database to S3

To upload a database to S3, use aws s3 cp. You can skip this section if you already have a kdb+ partitioned database in a object storage.

In our example, we will create a simple data set with randomized data.

// Generate data for today and the past few days last few days
n:1000000;
d:asc .z.d - til 3;
{[d;n]sv[`;.Q.par[`:data/db/;d;`trade],`]set .Q.en[`:data/;([]sym:`$'n?.Q.A;time:("p"$d)+til n;price:n?100f;size:n?50f)];}[;n] each d;

Now we will upload this example data.

aws s3 cp --recursive "data" "s3://insights-example-data/"

The sym file above top of the database directory is an enumeration of all symbols in the table. Under the trade table, there is a sym column which will reference the top-level file with the indices of relevant symbol names.

data
├── db
│   ├── 2023.05.09
│   │   └── trade
│   │       ├── price
│   │       ├── size
│   │       ├── sym
│   │       └── time
│   ├── 2023.05.10
│   │   └── trade
│   │       ├── price
│   │       ├── size
│   │       ├── sym
│   │       └── time
│   └── 2023.05.11
│       └── trade
│           ├── price
│           ├── size
│           ├── sym
│           └── time
└── sym

Additionally, a par.txt needs to be added in a different location than the database. In this example, the par.txt file will contain the following content.

par.txt

s3://insights-example-data/data/db

aws s3 cp par.txt s3://insights-example-data/data/par.txt

This par.txt is used below when mounting the database.

File locations

The sym and par.txt file must be in a different folder than the actual partitioned data. If they are in the same file, deploying the database will result in a 'part error as kdb+ is unable to mount the partitioned database.

(Optional) Creating a service account

If you are running in Kubernetes, you can use a service account to allow read access to AWS S3 without using environment variables. This can be done using eksctl.

eksctl create iamserviceaccount --name kx-s3-read-access \
    --namespace <your namespace> \
    --region <your region> \
    --cluster <your cluster> \
    --attach-policy-arn arn:aws:iam::aws:policy/AmazonS3ReadOnlyAccess \
    --approve

Deploying

The static HDB uses a segmented database that points to the object storage bucket which is configured with a par.txt file. Additionally, to maintain performance with symbol queries, we need to directly mount the sym file into the database.

The configuration to drive the object storage tier uses a microservice assembly. This assembly has a single object mount with an explicit sym and par configuration on the DAP instance which point to the sym and par.txt generated previously.

asm.yaml

name: odb
labels:
  name: odb-example
  type: odb
tables:
  trade:
    type: partitioned
    prtnCol: time
    columns:
      - name: time
        type: timestamp
      - name: sym
        type: symbol
      - name: price
        type: float
      - name: size
        type: float
mounts:
  odb:
    type: object
    baseURI: file:///data/db/odb
    partition: none
elements:
  dap:
    instances:
      odb:
        mountName: odb
        sym: s3://insights-example-data/data/sym
        par: s3://insights-example-data/data/par.txt

DockerKubernetes

This example deploys the object storage tier using Docker Compose to orchestrate the deployment.

Prerequisites

This example uses a .env file that will be specific to your deployment when you set up your environment and is not defined in this example. This example below uses the following environment variables.

variable	description
kxi_da	This is the URL to the kxi-da image. This should be the full image and tag URL. For example, portal.dl.kx.com/kxi-da:1.2.3
AWS_REGION	The region of the AWS bucket to query data from.
AWS_ACCESS_KEY_ID	Your AWS access key ID to programmatically access the specified bucket.
AWS_SECRET_ACCESS_KEY	Your AWS secret access key to programmatically access the specified bucket.

docker-compose.yaml

services:
  dap:
    image: ${kxi_da}
    env_file: .env
    environment:
      - KXI_SC=odb
      - KXI_ASSEMBLY_FILE=/data/asm.yaml
      - AWS_REGION=${AWS_REGION}
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - KDB_LICENSE_B64=${KDB_LICENSE_B64}
    command: ["-p", "5080"]
    ports:
      - 5080:5080
    volumes:
      - ./odb:/data/db/odb
      - ./asm.yaml:/data/asm.yaml

Writable volume

When running this example, a local directory odb will be created. This directory needs to be writable for the DAP to download the sym and par.txt configuration.

Now start the DAP.

docker compose up

This will start the DAP and present the data from object storage. This example can be combined with other deployment examples to leverage all query APIs.

In the example below, an object storage tier database is deployed into Kubernetes as a pod configuration. A persistent volume claim is used to hold a local cache of the sym and par.txt files from the object storage bucket. A config map is used to hold the microservice assembly configuration.

deploy.yaml

---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: odb-pvc
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 10Gi
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: odb-cm
data:
  asm.yaml: |
    name: odb
    labels:
      name: odb-example
      type: odb
    tables:
      trade:
        type: partitioned
        prtnCol: time
        columns:
          - name: time
            type: timestamp
          - name: sym
            type: symbol
          - name: price
            type: float
          - name: size
            type: float
    mounts:
      odb:
        type: object
        baseURI: file:///data/db/odb
        partition: none
    elements:
      dap:
        instances:
          odb:
            mountName: odb
            sym: s3://insights-example-data/data/sym
            par: s3://insights-example-data/data/par.txt
---
apiVersion: v1
kind: Pod
metadata:
  name: odb
spec:
  containers:
  - name: odb
    image: portal.dl.kx.com/kxi-da:1.5.0
    ports:
    - containerPort: 5080
    env:
      - name: KXI_SC
        value: odb
      - name: KXI_ASSEMBLY_FILE
        value: /cfg/asm.yaml
      - name: AWS_REGION
        value: "us-east-2"
      - name: AWS_ACCESS_KEY_ID
        value: "[redacted]"
      - name: AWS_SECRET_ACCESS_KEY
        value: "[redacted]"
      - name: KDB_LICENSE_B64
        value: "[redacted]"
    volumeMounts:
      - name: config
        mountPath: /cfg
      - name: data
        mountPath: /data/db/odb
    args: ["-p", "5080"]
  volumes:
    - name: config
      configMap:
        name: odb-cm
    - name: data
      persistentVolumeClaim:
        claimName: odb-pvc
  securityContext:
    fsGroup: 65535

To deploy this example, run the following.

kubectl apply -f deploy.yaml

This will deploy the object storage DAP tier locally into a Kubernetes cluster.