postponed

An HTTP sidecar service, wrapping the great open source bull library, for easy to use job scheduling. Bull is stable, well docummented library, that uses Redis as an underlying DB to manage jobs and queues. The sidecar behaved as both the producer and the consumer of the defined job type/queue. Exposes prometheus metrics for monitoring and alerts

Motivation

• Need X to happen Y minutes from now
• Possibly long periods of time- an hour or more

Example:

Say we have a chat service, where, when a user receives a message, he needs to be reminded or notified to read it after 1 minute, if he hasn't done so yet. So we have a service, which listens to some kind of message queue, and upon each arriving message, it schedules a job to be executed in 1 minute. The job is scheduled via sending a request to the sidecar's POST endpoint, and scheduling a job with all the necesary data in it (the recepient's and sender's id and name). When the job's turn arrives, the sidecar sends a POST request to a predefined endpoint on the service, with all the job's data (that was send to the sidecar), and then the service can check if the user has already seen the message, and notify him if not.

How to install

The docker image is hosted on docker hub and can be found here: https://hub.docker.com/r/mishaled/postponed/

An example kubernetes replication controller can be found in the examples directory.

API

Exposes a simple api

• Schedule a Job

  • http://localhost:**HTTP_PORT**

    • POST

    • Will call the host back when the job is active on TARGET_BASE_PATH + TARGET_PROCESS_RELATIVE_PATH via a post request with the jobData as the body

    • Expected body

      {
          jobId?: string;
          jobData: Record<string, any>;
          delayInSec: number;
      }

• Remove a Job

  • http://localhost:**HTTP_PORT**/:id
    • DELETE
    • id is the id of the job we need to remove

Environment configuration variables

Parameter	Description	Default	Mandatory	Secret
REDIS_URL	The redis cache url, if IS_REDIS_TLS is set to true, the REDIS_URL must start with rediss://, otherwise with redis://		true	true
RETRY_DELAY_MS	Time to wait before retrying to add to cache	2000	false	false
HTTP_PORT	The port to use	3000	false	false
TARGET_BASE_PATH	Target base path	http://localhost:3000	false	false
TARGET_PROCESS_RELATIVE_PATH	Target message processor path	/executeJob	false	false
TARGET_IS_ALIVE_RELATIVE_PATH	Target is alive path	isAlive	false	false
JOBS_QUEUE_NAME	The table name in redis to store the items	jobs-queue	false	false
IS_REDIS_TLS	Should connect to redis with tls	false	false	false
MAX_CONCURRENCY	The maximal number of concurrent processes	5	false	false
PROCESS_TIMEOUT_MS	Time after which the process times out in miliseconds	10000	false	false
RETRY_ATTEMPTS	The maximal number of retries	3	false	false
SHOULD_REDACT_JOB_DATA_IN_LOGS	Should redact the job.data from the logs, can be disabled if the data is not sensitive (PII etc.)	true	false	false

Exposed Prometheus Metrics

Name	Type	Description
schedule_new_job_start	Counter	Incremented each time the process of scheduling a new job starts
schedule_new_job_success	Counter	Incremented each time the process of scheduling a new job succeeds
schedule_new_job_error	Counter	Incremented each time the process of scheduling a new job fails
schedule_new_job_execution_time	Histogram	Represents the time it takes to schedule a new job
remove_job_start	Counter	Incremented each time the process of removing a job starts
remove_job_success	Counter	Incremented each time the process of removing a job succeeds
remove_job_error	Counter	Incremented each time the process of removing a job fails
remove_job_execution_time	Histogram	Represents the time it takes to remove a job
process_job_start	Counter	Incremented each time the processing of a job starts
process_job_success	Counter	Incremented each time the processing of a job succeeds
process_job_error	Counter	Incremented each time the processing of a job fails
process_job_execution_time	Histogram	Represents the time it takes to process job