Downstream Resiliency

What is the purpose of setting timeouts for network calls?

• To reduce resource leaks

• To propagate faults to other components

• To increase the response time

• To improve network connectivity

Which network API does not have a default timeout?

• JavaScript's XMLHttpRequest

• Python's requests library

• Go's HTTP package

• .NET Core HttpClient

How can we determine a good timeout duration?

• Based on the desired false timeout rate

• By setting it to infinity

• By measuring the response time of the caller

• By using the default timeout provided by the API

What should be measured at the integration points of systems?

• The duration of the network call

• The status code received

• Whether a timeout was triggered

• All of the above

What is the ideal way to handle network calls?

• Wrap them within a library function that sets a timeout and monitors the request

• Set a timeout for each call individually

• Ignore setting timeouts and rely on default values

• Avoid making network calls altogether

What are the two options a client has when a network request fails or times out?

• Fail fast or retry the request

• Increase the timeout or ignore the failure

• Switch to a different network protocol or cancel the request

• Delegate the request to a reverse proxy or use a sidecar pattern

What is the formula for calculating the delay between retries using exponential backoff?

• delay = min(cap, initial-backoff. 2attempt)

• delay = initial-backoff + attempt

• delay = cap / (initial-backoff * attempt)

• delay = initial-backoff - attempt

What is the purpose of introducing random jitter in the delay calculation for retries?

• To spread retries out over time and avoid herding behavior

• To increase the delay between retries exponentially

• To synchronize retries across multiple clients

• To reduce the overall delay between retries

In batch applications, how can failed network requests be retried?

• By parking the failed requests in a retry queue and retrying them later

• By immediately retrying the failed requests

• By cancelling the failed requests and ignoring them

• By delegating the retry process to a separate service

What is the potential problem with having retries at multiple levels of the dependency chain?

• Retry amplification and increased likelihood of timeouts

• Increased resilience and improved performance

• Reduced latency and faster response times

• Improved fault tolerance and error handling