Cicada scheduler

Centralized Distributed Scheduler

• Overview
  • Advantages over using only CRON
  • Execution flow diagram
• Setup Central database
  • Database ERD
• Setup Node
• Administration
  • cli help
  • (Re)register a server to Cicada
  • Add or Update a new schedule
  • Launch a schedule once
  • Launch a schedule once
  • Abort a running schedule

Overview

• Cicada is best understood as a lightweight management layer between linux CRON and the end executables
• It requires a central database for setting up and monitoring schedules
• Schedules can be launched on any node that has registered itself to the database
• Nodes act as agents pulling scheduling information from the database.

Advantages over using only CRON

• A central log is kept of schedule executions and results
• Offers some schedule dependance management
• Non-running schedules can easily be moved from one node to another
• Running schedules can easily be terminated

Execution flow diagram

Setup Central Database

Verified on PostgreSQL versions 12.9 to 15.14

1. Execute as postgres user setup/db_and_user.sql
2. Change cicada user password
3. Execute as cicada user setup/schema.sql

Database ERD

Editable Excalidraw source

Setup Node

Verified on Ubuntu 18.04, 20.04 LTS and 22.04 LTS

Prerequisites

• ntpd
• cron
• python3.8

# Install Cicada Scheduler
sudo mkdir -p /opt/cicada
sudo chown -R $USER:$USER /opt/cicada
cd /opt/cicada
DIR=$(pwd)

cd $DIR
git clone git@github.com:transferwise/cicada.git .
make python=python3.8

# Update the db_cicada section of the environmental config file
cp $DIR/config/example.yml $DIR/config/definitions.yml
vim $DIR/config/definitions.yml

# Register new Node in Database
$DIR/bin/cicada register_server

# Add linux CRON job to check central scheduler every minute
echo "* * * * * $DIR/venv/bin/python3 $DIR/venv/bin/cicada exec_server_schedules" | crontab

Administration

cli help

cicada --help

(Re)register a server to Cicada

cicada register_server

Add or Update a new schedule

cicada upsert_schedule --help

schedule_id             | Id of the schedule
schedule_description    | Schedule description and comments
server_id               | Id of the server where the schedule will run
schedule_order          | run order for the schedule. lowest is first. is_async jobs will be execute in parallel
is_async                | 0=Disabled 1=Enabled | is_async jobs execute in parallel
is_enabled              | 0=Disabled 1=Enabled
adhoc_execute           | 0=Disabled 1=Enabled | Execute at next minute, regardless of other settings
abort_running           | 0=Disabled 1=Enabled | If the job is running, it will be terminated as soon as possible
interval_mask           | When to execute the command | unix crontab (minute hour dom month dow)
first_run_date          | The schedule will not execute before this datetime
last_run_date           | The schedule will not execute after this datetime
exec_command            | Command to execute
parameters              | Parameters for exec_command
adhoc_parameters        | If specified, will override parameters for one run
schedule_group_id       | Optional field to help with schedule grouping

Launch a schedule once, as soon as possible

• adhoc_execute is a one-time schedule modification.
• As soon as adhoc_execution has been used, it is reset.

cicada upsert_schedule --schedule_id={schedule_id} --adhoc_execute={1}

Launch the next run of a schedule with different parameters

• adhoc_parameters is a one-time schedule modification.
• As soon as adhoc_parameters has been used, it is reset.

cicada upsert_schedule --schedule_id={schedule_id} ==adhoc_parameters={do something diffident}

Abort a running schedule

• abort_running is a one-time schedule modification.
• As soon as abort_running has been used, it is reset.
• Cicada keeps the schedule marked as running until the launched process exits.
• Additional abort requests are consumed while Cicada waits for the process to exit.

cicada upsert_schedule --schedule_id={schedule_id} --abort_running={1}