Terminus — Graceful Process Lifecycle Management

The world runs on processes. Most of them die badly.

Every meaningful computation eventually reaches its end. The question is whether that ending is orderly or catastrophic. For decades, the infrastructure community has treated process termination as an afterthought — a brute-force operation buried in system internals, wielded without nuance, understood by few. The result is orphaned state, corrupted data, and cascading failures at scale.

We founded Terminus on a first-principles conviction: the signal is the most powerful primitive in systems engineering. It is the mechanism through which software communicates intent across process boundaries. Graceful shutdown, forced resolution, handler verification, escalation pipelines — these are not edge cases. They are the foundation of reliable infrastructure. We are building the platform that treats them as such.

Process lifecycle management is not a feature. It is a category. And we are defining it.

Capabilities

Every signal. Every process.
Total control.

Terminus surfaces the full depth of modern signal infrastructure through a composable, auditable platform.

◇

Graceful Teardown Protocol™

Our default delivery mechanism sends TERM — the gold standard for orderly shutdown. Processes receive the signal, execute cleanup handlers, and exit with dignity. No data loss, no orphaned state.

--signal TERM

⬡

Force Resolve™

When graceful is not an option, escalate to KILL — an uncatchable, unblockable termination vector. The process does not get a say. For when reliability demands certainty over courtesy.

-9

◈

Signal Dispatch Engine™

Specify any signal by name or number. HUP, INT, QUIT, USR1, USR2, CONT, STOP — the full POSIX spectrum, dispatched with sub-millisecond precision to any target process.

-s SIGNAL

△

Escalation Pipeline™

Define multi-stage termination workflows with configurable timeout intervals. Send QUIT, wait, then TERM, wait, then KILL — a sequential resolution strategy executed atomically, immune to PID recycling.

--timeout ms SIGNAL

○

Connection Probe™

Validate process reachability without disruption. Signal 0 performs full permission and existence checks with zero side effects — the ideal health-check primitive for monitoring infrastructure.

-s 0

▽

Signal Intelligence Dashboard™

Enumerate the complete signal namespace. Convert between names, numbers, and bitmasks. Full introspection into the signal taxonomy your system supports.

-l, -L

◎

Handler Verification™

Only deliver signals to processes that have explicitly registered a userspace handler. Eliminate wasted signals and unintended terminations with require-handler mode.

--require-handler

⬢

Signal Queue Delivery™

Attach integer payloads to signals via the queue interface. Receiving processes access the data through SA_SIGINFO handlers — structured inter-process communication at the kernel level.

-q value

Architecture

From intent to resolution
in four stages

Identify

Target processes by PID, process name, or group. Terminus resolves identifiers across your entire process tree with cross-UID support for full-system visibility.

Configure Signal

Select from the complete signal spectrum — by name or number. Attach queue payloads, set escalation timeouts, or enable handler verification. Define your intent precisely.

Dispatch

Terminus delivers the signal to the kernel with sub-millisecond latency. For escalation pipelines, subsequent signals are dispatched via PID file descriptors — race-free and atomic.

Verify

Confirm delivery status, inspect exit codes, and audit the complete signal lifecycle. Verbose mode provides full transparency into every PID that received your signal.

In Action

See Terminus at work

Real signal delivery. Real process management. Select a scenario to observe the platform in action.

terminus — signal dispatch

$ terminus 8421
# Sends SIGTERM (15) to process 8421
$ echo $?
0

$ terminus -9 8421
# Sends SIGKILL (9) — uncatchable, immediate
$ echo $?
0

$ terminus --verbose --timeout 1000 TERM --timeout
                            1000 KILL \
  --signal QUIT 12345
sending signal SIGQUIT to pid 12345
sending signal SIGTERM to pid 12345
sending signal SIGKILL to pid 12345

$ terminus -s 0 8421
# Signal 0: existence check, no signal sent
$ echo $?
0
$ terminus -s 0 99999
terminus: sending signal to 99999 failed: No such process
$ echo $?
1

Case Study

How NovaCast eliminated orphaned processes at scale

NovaCast operates a distributed media transcoding pipeline across 14,000 compute nodes. Before Terminus, their engineering team relied on ad-hoc signal delivery to manage worker processes — a fragile approach that left orphaned processes consuming resources for hours after jobs completed. The cost: an estimated 23% wasted compute spend and frequent capacity-related outages during peak traffic.

After adopting Terminus, NovaCast implemented structured Escalation Pipelines for all worker lifecycle management. Each transcoding worker now receives a QUIT signal, followed by TERM after 500ms, and KILL after 2000ms — a three-stage teardown that guarantees cleanup while ensuring no process survives past its useful life. The team also deployed Connection Probe for health monitoring, reducing false-positive alerts by 89%.

"We went from firefighting orphaned processes every morning to not thinking about process lifecycle at all. Terminus turned a daily operational burden into a solved problem. The Escalation Pipeline alone paid for the entire contract in the first week."

— Jordan Hale, VP of Platform Engineering, NovaCast

Pricing

Simple, transparent pricing

Start free. Scale when you need to.

Starter

$0/mo

SIGTERM only
100 signals / month
Single PID targeting
Basic exit code reporting
Signal list / table
Escalation Pipelines
Queue delivery

Get Started

Trusted by infrastructure teams worldwide

"We replaced an entire internal tool with Terminus. The Escalation Pipeline alone saved us from writing 600 lines of bash. This is what infra should feel like — invisible and correct."

David Fischer SRE Lead, CloudForge

"Signal 0 as a health check primitive is criminally underrated. Terminus made it first-class. Our monitoring false positives dropped to near zero. Genuinely game-changing."

Ethan Sato VP of Infrastructure, ScaleForge

"I was sending SIGKILL to everything like a barbarian. Terminus taught my team about graceful teardowns. Our data corruption incidents went from weekly to zero. Not an exaggeration."

Derek Moss CTO, DataMesh

"The queue delivery feature with SA_SIGINFO is the most elegant IPC mechanism available. Terminus wraps it in an interface that doesn't require reading kernel docs. That is a 10x unlock."

Owen Stollenberg Solo Developer & Systems Consultant

FAQ

Questions we hear often

What happens if the process doesn't terminate after receiving a signal? +

Some signals, like TERM, can be caught by the process — meaning the process may choose to handle it and continue running. This is by design: it allows orderly cleanup. If termination is required regardless, use Force Resolve (SIGKILL, signal 9), which cannot be caught, blocked, or ignored. For automated workflows, our Escalation Pipeline feature sends a sequence of signals with configurable timeouts, ensuring eventual resolution.

Is my data secure during signal dispatch? +

Terminus operates at the kernel level with strict permission enforcement. You can only signal processes you own, unless operating with elevated privileges. Signal 0 lets you probe process existence without any side effects. All dispatch operations are atomic — there is no intermediate state where data could be compromised. We are SOC 2 Type II certified and GDPR compliant.

Can I send signals to multiple processes at once? +

Yes. Terminus supports both individual PID targeting and process group dispatch. Use a negative PID to signal an entire process group atomically. You can also target processes by name, and with the cross-UID flag, extend your reach across user boundaries for full system-level management.

What if I accidentally send the wrong signal? +

Terminus supports Handler Verification mode, which checks that the target process has a registered userspace handler for the signal before delivery. Combined with Dry Run mode (--pid flag), you can audit exactly which processes would receive a signal without actually sending it. For SIGCONT and similar non-destructive signals, accidental dispatch has no adverse effect. For destructive signals, we strongly recommend Escalation Pipelines — start gentle, escalate only when needed.

Does Terminus work offline? +

Terminus operates entirely at the system level with zero network dependencies. Signal dispatch is a local kernel operation — it functions identically whether your machine is connected to the internet or completely air-gapped. Our Enterprise tier includes on-premise deployment for regulated environments that require full network isolation.

Terminate withprecision.