Database drivers
The Stackable product images for Apache Hive come with built-in support for using PostgreSQL as the metastore database. The MySQL driver is not shipped in our images due to licensing issues. To use another supported database it is necessary to make the relevant drivers available to Hive: this tutorial shows how this is done for MySQL.
Install the MySQL helm chart
helm install mysql oci://registry-1.docker.io/bitnamicharts/mysql \
--set auth.database=hive \
--set auth.username=hive \
--set auth.password=hive
Download the driver to a PersistentVolumeClaim
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: pvc-hive-drivers
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 1Gi
Download the driver from e.g. maven to a volume backed by the PVC:
---
apiVersion: batch/v1
kind: Job
metadata:
name: pvc-hive-job
spec:
template:
spec:
restartPolicy: Never
volumes:
- name: external-drivers
persistentVolumeClaim:
claimName: pvc-hive-drivers
initContainers:
- name: dest-dir
image: docker.stackable.tech/stackable/tools:1.0.0-stackable24.3.0
env:
- name: DEST_DIR
value: "/stackable/externals"
command:
[
"bash",
"-x",
"-c",
"mkdir -p ${DEST_DIR} && chown stackable:stackable ${DEST_DIR} && chmod -R a=,u=rwX,g=rwX ${DEST_DIR}",
]
securityContext:
runAsUser: 0
volumeMounts:
- name: external-drivers
mountPath: /stackable/externals
containers:
- name: hive-driver
image: docker.stackable.tech/stackable/tools:1.0.0-stackable24.3.0
env:
- name: DEST_DIR
value: "/stackable/externals"
command:
[
"bash",
"-x",
"-c",
"curl -L https://repo1.maven.org/maven2/com/mysql/mysql-connector-j/8.0.31/mysql-connector-j-8.0.31.jar \
-o ${DEST_DIR}/mysql-connector-j-8.0.31.jar",
]
volumeMounts:
- name: external-drivers
mountPath: /stackable/externals
This will make the driver available at /stackable/external-drivers/mysql-connector-j-8.0.31.jar
when the volume external-drivers
is mounted at /stackable/external-drivers
.
Once the above has completed successfully, you can confirm that the driver is in the expected location by running another job:
---
apiVersion: batch/v1
kind: Job
metadata:
name: list-drivers-job
spec:
template:
spec:
restartPolicy: Never
volumes:
- name: external-drivers
persistentVolumeClaim:
claimName: pvc-hive-drivers
containers:
- name: hive-driver
image: docker.stackable.tech/stackable/tools:1.0.0-stackable24.3.0
env:
- name: DEST_DIR
value: "/stackable/externals"
command:
[
"bash",
"-x",
"-o",
"pipefail",
"-c",
"stat ${DEST_DIR}/mysql-connector-j-8.0.31.jar",
]
volumeMounts:
- name: external-drivers
mountPath: /stackable/externals
Create a Hive cluster
The MySQL connection details can then be used in the definition of the Hive Metastore resource.
Note that it is also necessary to "tell" Hive where to find the driver.
This is done by setting the value of the environment variable METASTORE_AUX_JARS_PATH
to the path of the mounted driver:
---
apiVersion: hive.stackable.tech/v1alpha1
kind: HiveCluster
metadata:
name: hive-with-drivers
spec:
image:
productVersion: 3.1.3
clusterConfig:
database:
connString: jdbc:mysql://mysql:3306/hive (1)
user: hive (2)
password: hive
dbType: mysql
s3:
reference: minio (3)
metastore:
roleGroups:
default:
envOverrides:
METASTORE_AUX_JARS_PATH: "/stackable/external-drivers/mysql-connector-j-8.0.31.jar" (4)
podOverrides: (5)
spec:
containers:
- name: hive
volumeMounts:
- name: external-drivers
mountPath: /stackable/external-drivers
volumes:
- name: external-drivers
persistentVolumeClaim:
claimName: pvc-hive-drivers
replicas: 1
1 | The database connection details matching those given when deploying the MySQL Helm chart |
2 | Plain-text Hive credentials will be replaced in an upcoming release! |
3 | A reference to the file store using S3 (this has been omitted from this article for the sake of brevity, but is described in e.g. the First steps guide) |
4 | Use envOverrides to set the driver path |
5 | Use podOverrides to mount the driver |
This has been tested on Azure AKS and Amazon EKS, both running Kubernetes 1.29.
The example shows a PVC mounted with the access mode ReadWriteOnce as we have a single metastore instance that is deployed only once the jobs have completed, and, so long as these all run after each other, they can be deployed to different nodes.
Different scenarios may require a different access mode, the availability of which is dependent on the type of cluster in use.
|
Alternative: using a custom image
If you have access to a registry to store custom images, another approach is to build a custom image on top of a Stackable product image and "bake" the driver into it directly:
FROM docker.stackable.tech/stackable/hive:3.1.3-stackable0.0.0-dev
RUN curl --fail -L https://repo1.maven.org/maven2/com/mysql/mysql-connector-j/8.0.31/mysql-connector-j-8.0.31.jar -o /stackable/mysql-connector-j-8.0.31.jar
docker build -f ./Dockerfile -t docker.stackable.tech/stackable/hive:3.1.3-stackable0.0.0-dev-mysql .
---
apiVersion: hive.stackable.tech/v1alpha1
kind: HiveCluster
metadata:
name: hive
spec:
image:
custom: docker.stackable.tech/stackable/hive:3.1.3-stackable0.0.0-dev-mysql (1)
productVersion: 3.1.3
clusterConfig:
database:
...
s3:
...
metastore:
config:
logging:
enableVectorAgent: False
roleGroups:
default:
envOverrides:
METASTORE_AUX_JARS_PATH: "/stackable/mysql-connector-j-8.0.31.jar" (2)
replicas: 1
1 | Name of the custom image containing the driver |
2 | Path to the driver |