Circuit Breaker Pattern

• It solve the following problems 
• Non-transient failures
• Resilient systems
• Cascading failures
• Fault tolerance
• Retries and recovery
• It is a cloud design pattern To deal with non-transient failure in micro services.
• A non-transient failure is a permanent failure that can render the system Unavailable or the recovery can take longer than few seconds.
• Examples of non-transient failure are
• Timeout
• Database failure
• Connection failure
• Services bombarded with multiple request and all the request resulting in failure due to memory or CPU issue in application.
• Circuit breaker helps to Avoid cascading failures, where downtime of one service impacts other dependent services.
• We try to stop the calls to the failing service until it recovers.
• The circuit breaker Logic will decide if the request from A will be passed through or rejected or returned with an error.
• There are three states of circuit breaker
• Closed
• When closed all traffic goes as normal.
• Half open
• When half open some traffic Go for testing purposes, if the calling service can call downstream service and check if there are any failure or not.
• Open
• When the circuit breaker is open The calling service will not be able to call downstream service and will return an error message.
• When half open if the service recovers, the circuit goes into close, state, else it goes into open state.
• Default state is closed.
• After failure, the circuit keeps switching between open and half open state to check recovery.
• We can make our circuit pattern more efficient by having a control over type of and count of errors.
• We should also have manual control Over states, for example, if we want to open the circuit breaker, even though services are doing well.
• We should keep a check on timeout for different states.
• We should check for concurrency as one service might be getting called from many other services.
• We should log and monitor different states. 
• We do not need This, when we have a few services or scenarios, where load is not heavy.
• Effective way to bring resilience in the systems.
• Should not be used to solve transient failures.