job index backoff limit

gistfile1.txt

client sends this
backoffLimit: nil

client gets back this
backoffLimit: 6


backoffLimit: xxx
runPolicy:
  perIndex: true/false


new server has this stored and then downgrades
backoffLimit: 10
runPolicy:
  perIndex: true
  
on downgrade you end up reading
backoffLimit: 10

in this scenario there is nothing we can recommend.




new server has this stored and then downgrades
backoffLimit: nil
perIndexBackoffLimit: 10
  
on downgrade you end up reading
backoffLimit: 6

what would we tell someone who wants to span versions, we can recommend something like
backoffLimit: 100 // perIndexBackoffLimit * expectedFailureRate * completionRuns
perIndexBackoffLimit: 10
maxIndexFailures: 10


index:   1   2   3   4   5
fails:   5   4   7   2   4


new client to new server create
backoffLimit: nil
perIndexBackoffLimit: 10

new client gets back
backoffLimit: nil
perIndexBackoffLimit: 10



imagine BEFORE backoffLimit exist, we have this stored

on read after backoffLimit exists
backoffLimit: 6