Resilience and failure handling

you give up on the call after two seconds instead of holding the thread open until the other side eventually answers or does not.

Timeout

you reject the request immediately when you already know the dependency is down, instead of making the caller wait for a timeout.

Fail fast

you pass the remaining time budget down with the request so a service four hops deep knows there is no point starting.

Deadline propagation

the dependency got slow, everything retried, and the extra load is now the reason it cannot recover.

Retry storm

you add a random amount to every retry delay so ten thousand clients do not all come back at exactly the same second.

Jitter

you cap retries at a small fraction of total traffic so a struggling dependency does not get three times the load it was already failing under.

Retry budget

the call timed out and you genuinely do not know whether the other side did the work or not.

Partial failure

one service got slow, its callers filled up with waiting requests, and the failure walked all the way up to the front door.

Cascading failure

you ask how much of the product breaks if this one component dies, and design so the answer is not "all of it".

Blast radius

you trace the dependency graph and find one component whose failure takes everything else down with it.

Single point of failure

you give each downstream its own pool of connections so one slow dependency cannot consume every thread you have.

Bulkhead

the consumer tells the producer to slow down rather than silently buffering work it cannot keep up with.

Backpressure

you start rejecting requests immediately with a 503 instead of accepting work you have no chance of finishing in time.

Load shedding

the permissions service timed out and you had to decide, in advance, whether to let the request through or block it.

Fail open

the orchestrator asks whether the process is wedged and restarts the container when it stops answering.

Liveness probe

the instance says "not yet" while the cache warms, so the load balancer keeps traffic off it until it is genuinely able to serve.

Readiness probe

on SIGTERM you stop accepting new requests, finish the ones in flight, and only then exit.

Graceful shutdown

you deliberately kill an instance in production during working hours to find out whether the failover actually works.

Chaos engineering