Feature request: make default maxFailedTasks 0

[crowecawcaw]

Describe the problem

Current the default value for maxFailedTasks is 5 which is gratuitous. In all likelihood, a job that fails on its first try will fail on subsequent tries. The extra attempts slow iterations if we're waiting for a job to succeed or fail.

Proposed Solution

Make the default 0. If someone has a job that is not reliable, they can opt in to retries.

Example Use Cases

...

[karthikbekalp]

Hmm, I assume here you mean the maxRetriesPerTask and not maxFailedTasksCount?

I agree with you about maxRetriesPerTask and if individual DCCs need multiple retries, they can opt in to retries.

But I'll discuss with SMEs and get back to you on this. :)

[karthikbekalp]

I spoke about this internally with our SMEs, and it looks like these defaults were carefully selected after speaking with our internal teams.

Turns out that though 0 retries is great for development/debugging, for production workloads it's terrible and instead causes more customer pain.

Let me add this to the docs somewhere so that we know why we made these decisions.

[karthikbekalp]

Hmm, looks like we might have newer data that might change our decision.

We will discuss this internally within our team and update here.