rsched

= rsched

A Generic Reliable Scheduler. It's like cron, but rsched supports redundancy using multiple servers.

== Architecture

1. rsched virtually locks a record on a RDBMS.
2. if the schedule is not finished and not locked by other node, run it.
3. if it succeeded, mark it finished.
4. if it failed, unlock it and expect to be retried.

== Install

$ gem install rsched $ gem install dbd-mysql # to use MySQL as a lock database $ gem install dbd-sqlite3 # to use SQLite3 as a lock database

== Schedule

A schedule consists of ident, time and action. Ident describes unique identifier of the schedule that. Time describes when it should be scheduled. Action is an description of the schedule.

Format of the time is same as cron. See man 5 cron for details.

Example:

Run every minutes

$ rsched -a 'mysched * * * * * my descriptoin of this action' ...

Run every day at 00:00

$ rsched -a 'mywork 0 0 * * * aaaa uuu ee' ...

Use SQLite3 for the lock database

$ rsched -T mysql -H myhost -D mydb -u myuser -p mypassword ...

Use MySQL for the lock database

$ rsched -T sqlite3 -D /path/to/db.sqlite3 ...

== Usage

Usage: rsched [options] [-- ] --configure PATH.yaml Write configuration file --exec COMMAND Execute command --run SCRIPT.rb Run method named 'run' defined in the script -a, --add EXPR Add an execution schedule -t, --timeout SEC Retry timeout (default: 30) -r, --resume SEC Limit time to resume tasks (default: 3600) -E, --delete SEC Limit time to delete tasks (default: 2592000) -n, --name NAME Unique name of this node (default: PID.HOSTNAME) -w, --delay SEC Delay time before running a task (default: 0) -F, --from YYYY-mm-dd_OR_now Time to start scheduling -e, --extend-timeout SEC Threashold time before extending visibility timeout (default: timeout * 3/4) -x, --kill-timeout SEC Threashold time before killing process (default: timeout * 10) -X, --kill-retry SEC Threashold time before retrying killing process (default: 60) -i, --interval SEC Scheduling interval (default: 10) -U, --release-on-fail Releases lock if task failed so that other node can retry immediately -T, --type TYPE Lock database type (default: mysql) -D, --database DB Database name -H, --host HOST[:PORT] Database host -u, --user NAME Database user name -p, --password PASSWORD Database password --env K=V Set environment variable -d, --daemon PIDFILE Daemonize (default: foreground) -f, --file PATH.yaml Read configuration file

Lock database (-T) is used to synchronize scheduling status over multiple servers. Rsched supports following database types:

• mysql uses MySQL as a lock database. Note that 'dbd-mysql' gem must be installed.
• sqlite3 uses SQLite3 as a lock database. Note that 'dbd-sqlite3' gem must be installed.

One of --exec, --run or --configure is required. The behavior of the commands is described below:

=== exec

Execute a command when an action is scheduled. ident, time and action is passed to the stdin with tab-separated format. The command have to exit with status code 0 when it succeeded.

Example:

#!/usr/bin/env ruby ident, time, action = STDIN.read.split("\t", 3) t = Time.at(time.to_i) puts "scheduled on #{t} ident=#{ident} action=#{action}"

$ rsched -a 'mysched * * * * * des' -T sqlite3 -D test.db -F now --exec ./this_file

=== run

This is same as 'exec' except that this calls a method named 'run' defined in the file instead of executing the file. It is assumed it succeeded if the method doesn't any raise errors.

Example:

def run(ident, time, action) t = Time.at(time) puts "scheduled on #{t} ident=#{ident} action=#{action}" end

$ rsched -a 'mysched * * * * * des' -T sqlite3 -D test.db -F now --exec ./this_file.rb