A Simple python based scheduler to schedule kubernetes jobs on crontab schedules.
This module will help schedule two kinds of jobs ,
These are jobs which are supposed to run on a crontab schedule. This package would launch kubernetes pods as per the crontab schedule and monitor its runs.
These are jobs which are supposed to be running continuously. Like a stream data input. This package would launch and monitor such continuous runs. If any of the run fails for any reason , this package would reschedule it.
To install the scheduler run below command,
pip install k8s-scheduler
As the name suggests it's a job scheduler for kubernetes , so you would need a k8s instance apart from below mandatory and optional requirements.
Mongodb - This acts as a data store for scheduler and results backend.
EFL / ELK Stack - For logging and retrieving logs for a pod which is complete / running
Use the package to create a schedule entry in the backend , and use the run_scheduler method to start the scheduler in backrgound mode.
from k8s_scheduler import Scheduler
from k8s_scheduler.LogHandler import LogHandler
scheduler = Scheduler(host="mongo_db_host",
username="user_name",
password="password",
db_name="test_schedule")
# Create schedules in the mongodb collection
job_id_1 = scheduler.create_schedule(schedule=test_job_1)
job_id_2 = scheduler.create_schedule(schedule=test_job_2)
job_id_3 = scheduler.create_schedule(schedule=test_job_3)
# Start Scheduler application
scheduler.start_scheduler()Below is the template for the scheduler object.
schedule_rec = {
"schedule_name": "test-schedule-3-always", # Name of the schedule
"schedule_enabled": 1, # 1 for enabled 0 for disabled
"schedule_description": "test always schedule with nginx and with service",
"schedule_type": "always", # always is for streaming continuous jobs , periodic is for crontab based schedules
"parallel_execution": 0, # If two instances of the same job can run together
"schedule_crontab": { # Crontab schedule
"minute": "*/2",
"hour": "*",
"day_of_month": "*",
"month": "*",
"day_of_week": "*"
},
"kubernetes_deployment_options": {
"name": "nginx-always-service", # name of the deployment
"container_name": "nginx-always", # Container name
"container_image": "nginx:latest", # Image
"restart_policy": "", # k8s Restart policy
"ttl_seconds_after_finished": 10, # Seconds until the job needs to be purged
"env_vars": [], # Env vars for the deployment
"deploy_service": 1, # If a service needs to be deployed ? 1=yes 0=No
"service_name": "nginx-service", # Name of the service
"port": 8080, # Port exposed
"target_port": 80 # Port exposed
}
}Below are the defaults when starting the scheduler, any of these params can be modified when creating a scheduler instance
scheduler = Scheduler(db_name="k8s_scheduler", # Name of the Mongodb Database
scheduler_collection_name="test_coll", # Collection name for the scheduler
result_db_collection='result_coll', # Collection name for the result store
scheduler_poll_interval=10, # Polling interval for scheduler in seconds
k8s_worker_namespace="test-namespace", # k8s namespace in which pods are to be deployed
k8s_config_file_path='/var/snap/microk8s/current/credentials/client.config', # k8s config file if running outside cluster
**kwargs # connection args for mongodb
)We can retrieve logs of the jobs by connecting to ELK/EFK stack as below
log_handler = LogHandler(elastic_search_server='10.1.179.109', # Elastic server
elastic_search_port='9200', # port
worker_namespace='test-namespace', # namespace of the workers
index='logstash*') # logstash index
log_handler.get_logs(schedule_name="nginx-always", # Name of the schedule
output_json=True) # If True Returns log as json object , if False prints log to console
To test the application we can use microk8s to spin up a k8s cluster and install required backends.
Install microk8s / k8s flavour
sudo snap install microk8s --classic
microk8s enable helm3 ingress dashboard dns storage registry fluentd
sudo snap alias microk8s.kubectl kubectl
sudo snap alias microk8s.helm3 helm
Install backends
helm install test-mongodb bitnami/mongodb -n test-mongodb --create-namespace --set architecture=replicaset
Once the stack is up use make file to create and clean the dns entries so that you can connect to database from outside the cluster.