middleware-stack

middleware-stack

Step and enqueue middleware on the in-memory driver: the order middleware.Chain runs in, Logging with a level per attempt (WithLogLevel), Recover, Timeout, Skip, ThrottleErrors with a CircuitStore of the app's own, FailOn, then WrapClient with an enqueue middleware of the app's own, Route and Unique, and the refusal when Unique is put in one chain twice.

cd examples
go run ./middleware-stack

The log lines carry no time or run ID and durations read 0s (Logging measures on a fake clock), so the output is the same on every run:

a before
b before
handler
b after
a after
-- step middleware
level=WARN msg="workflow step failed" workflow=charge step=run attempt=1 duration=0s permanent=false error="payments api down"
run charge: succeeded, 2 handler call(s)
level=ERROR msg="workflow step panicked" panic="nil map"
level=ERROR msg="workflow step failed" workflow=panics step=run attempt=1 duration=0s permanent=false error="workflow/middleware: step panicked: nil map"
run panics: failed (memory: step "run" failed after 1 attempt(s): workflow/middleware: step panicked: nil map), 1 handler call(s)
level=ERROR msg="workflow step failed" workflow=validate step=run attempt=1 duration=0s permanent=true error="field amount: bad input"
run validate: failed (memory: step "run" failed after 1 attempt(s): field amount: bad input), 1 handler call(s)
level=ERROR msg="workflow step failed" workflow=slow step=run attempt=1 duration=0s permanent=false error="context deadline exceeded"
run slow: failed (memory: step "run" failed after 1 attempt(s): context deadline exceeded), 1 handler call(s)
run charge: succeeded, 0 handler call(s)
-- throttle
level=ERROR msg="workflow step failed" workflow=call-api step=run attempt=1 duration=0s permanent=false error="payments api down"
run call-api: failed (memory: step "run" failed after 1 attempt(s): payments api down), 1 handler call(s)
circuit "call-api" open for 5m0s
level=ERROR msg="workflow step failed" workflow=call-api step=run attempt=1 duration=0s permanent=false error="payments api down"
run call-api: failed (memory: step "run" failed after 1 attempt(s): payments api down), 1 handler call(s)
level=ERROR msg="workflow step failed" workflow=call-api step=run attempt=1 duration=0s permanent=false error="workflow/middleware: step throttled after repeated errors (key \"call-api\")"
run call-api: failed (memory: step "run" failed after 1 attempt(s): workflow/middleware: step throttled after repeated errors (key "call-api")), 0 handler call(s)
level=ERROR msg="workflow step failed" workflow=call-api step=run attempt=1 duration=0s permanent=false error="payments api down"
run call-api: failed (memory: step "run" failed after 1 attempt(s): payments api down), 1 handler call(s)
-- enqueue middleware
enqueue charge tier="bulk" queue="bulk" err=<nil>
run charge: succeeded, 0 handler call(s)
enqueue charge tier="" queue="" err=workflow: duplicate idempotency key: unique key "o-7" is held
same order again: true
unique twice: true workflow: invalid spec: Unique: label "ubgo.workflow/unique-key" uses the reserved prefix "ubgo.workflow/"

What each part shows:

• charge fails once and is retried; WithLogLevel logs the retried attempt at Warn.
• panics panics; Recover turns it into an error wrapping middleware.ErrPanic, and Logging, outside it, logs the failed attempt.
• validate has two retries but FailOn makes its bad input error permanent, so it runs once.
• slow blocks on its context until Timeout ends it.
• A "skip" payload completes without a handler call.
• call-api fails twice, its circuit opens, the third run is throttled without a handler call, and after the cooldown the step runs again.
• Route adds queue=bulk to a tier=bulk job, Unique refuses a second job for the same order, and two Unique layers refuse every job with workflow.ErrInvalid.

The guides that copy this program: step middleware, enqueue middleware and logging. go test ./middleware-stack runs it and checks every line.