feat(KRY-624): A more intelligent SideKick

[kmushegi]

What it Does

• Use additional info from QuickSilver to make more intelligent decisions on traffic routing
  • if cluster is unhealthy, automatically try to hit S3
  • if cluster is healthy, split traffic between crunch and s3 based on the traffic splitting param
  • if cluster is healthy, and initial target is S3 and that fails with 404, fallback to Bolt
• Rename things all around since QuickSilver no longer just returning endpoints
• Add tests
• Update integration test instructions
• Add Python boto3 integration example

Integration tests are passing w/ sidekick url specified and without ✅

[jr-projectn]

if cluster is healthy, split traffic between crunch and s3 based on the traffic splitting param

is this read or write traffic. Are we handling read and write separately?
Is the fallback behavior included in this PR?

[kmushegi]

if cluster is healthy, split traffic between crunch and s3 based on the traffic splitting param

is this read or write traffic. Are we handling read and write separately? Is the fallback behavior included in this PR?

Some speculation on Traffic Split for writes, fallback for reads for an inline crunch workflow If traffic split > 50% -ping Bolt first for reads -fallback to S3 for writes -fallback to S3 for reads

read and write are not handled separately. fallback is already included in SideKick and is controlled via a documented arg.

In this iteration let's just split all requests independent of what they are. We can iterate on this down the line.

[jr-projectn]

+1