lacks falc job delete

[seamusabshere]

all you have to do is

1. delete the k8s job
2. set status=deleted in the jobs table

[seamusabshere]

@emk

[emk]

Yup, I saw this!

I'm not sure if we actually want "delete" or "cancel" semantics—I need to think a bit about the use cases here, and what produces desirable behavior.

[seamusabshere]

ah, you want stop instead of delete?

[emk]

Yeah, I'm not sure exactly what the most convenient design might be yet. For example, if a job is failing, it might be nice to cancel it but preserve enough debug info to investigate.

[seamusabshere]

here's what i do when a job dies unceremoniously and i need to get it ready for a retry:

kubectl job delete X
ruby cancel_job.rb X

cancel_job.rb:

require 'sequel'

db = Sequel.connect(`falconeri db url`.chomp)
db.transaction do
  sql = <<-SQL
    UPDATE
      datums
    SET
      status = 'error'
    WHERE
          status = 'running'
      AND job_id = (
        SELECT
          id
        FROM jobs
        WHERE
          job_name = #{db.literal ARGV[0]}
      )
  SQL
  warn "erroring datums..."
  warn sql
  db[sql].get
  warn "done"

  sql = <<-SQL
    UPDATE
      jobs
    SET
      status = 'error'
    WHERE
      job_name = #{db.literal ARGV[0]}
  SQL
  warn "erroring job..."
  warn sql
  db[sql].get
  warn "done"
end