

Have job dependencies.

Be able to run jobs remotely on hosts.

Be able to run all downstream jobs remotely (if necessary).

Be able to automatically retry jobs.

Have ability to create PROD/BACKUP from in the browser

Automatic notifications for job failures and successes.

Abililty to send reports. E.g. At 1900 Structured Data to get an email with the status of all jobs that we're meant to support,
modulo some logic about their cron schedule.

Put last run/succeeded on the jobs, and then have alerting around the cron schedule if it hasn't run in a specified
period of time.

Be able to distribute jobs amongst machines.

Automatically register jobs with Icinga.

Be able to log to write log files to the Tardis.

Have a way to assign all your support roles to another person/people for a period of time. E.g. when you're on vacation.
Ideally this would involve having a support roster, which is a time series database which maps people to jobs. Should a
job require support, Kaladin should look at the roster to determine who's on.
    Is there a way to have indefinite periodic support? E.g. A supports the job from 0500 to to 1700, and B from 1700 to 0500 T+1.
    Can probably do it by having valid_from, valid_to, and support_from and support_to.
    Or have a "meta-schedule" table, which updates the schedule table as necessary.


Be able to SMS/email users when their jobs require support

Have histories on tables

Be able to run job graphs with no schedules.


Be able to mark a node/job as terminal node. If a non-terminal doesn't feed something else, why not?