For DevOps and SRE
Uptime Monitoring for DevOps Teams
Sentinel waits for consensus from fifteen regions before anyone gets alerted. 15-second checks, PagerDuty, Terraform, and monitoring built for the person responsible for uptime.
Why DevOps teams pick Sentinel
Alerts you can act on at 3am
A single probe can mistake its own network problem for your outage. Sentinel checks from fifteen regions and only opens an incident when at least half of them agree, so an alert means the service is down, not that a route in Virginia hiccuped.
- Consensus across regions before anyone gets alerted
- Monitors as code with the Terraform provider and the CLI
- Deploy markers on every chart, straight from GitHub
Built for the on-call rotation
Everything an alert needs to be right
Multi-region consensus
Every check runs from fifteen regions on six continents. An incident needs at least half the regions, three or more on two continents, to agree, which is what keeps false alerts out of your on-call rotation.
API checks, not URL pings
Any method, your headers and body, basic, bearer or digest auth, the status codes you accept, and JSON assertions on the response. A 200 with the wrong payload is down.
PagerDuty, Slack, webhooks
Open and resolve PagerDuty incidents through the Events API, post to Slack, Discord or Teams, or send signed JSON to your own endpoint. Routing is per monitor, not per account.
Monitors as code
Declare monitors, groups and alert routing in Terraform or script them from the CLI. The same REST API sits under the dashboard, with write access on every plan, including Free.
Deploy markers
Install the GitHub app and every deployment is drawn on the response-time charts and cited on incidents, so the question 'what changed?' already has an answer.
Heartbeats for the jobs nobody watches
Cron jobs, queue workers and backups ping Sentinel when they finish. Miss a ping and you hear about it. Dead-man switches for the parts of the system that don't serve HTTP.
Outside in
Your dashboards say healthy. Your users disagree.
Datadog, Grafana and Prometheus tell you what the application sees. By the time a request shows up in a trace, DNS resolved, the handshake worked and the CDN let it through. The failures that don't make it that far never appear in your telemetry at all.
Sentinel checks from outside your infrastructure, from fifteen regions around the world, the way a user would. It sits next to your observability stack and pages the same on-call rotation when what users get and what the app reports come apart.
What your telemetry sees
- Request latency, measured once the request reaches the app
- Error rates, database time, queue depth, CPU and memory
- Traces of the code paths that ran
- Health checks that pass because the process is up
What Sentinel sees
- DNS resolving to the right place, from fifteen regions
- The TLS handshake and the certificate users actually get
- The CDN or WAF answering with its own error page instead of yours
- One continent timing out while the origin reports healthy
- Time to first byte as a user sees it, not as the app measures it
Monitoring as code
The repo is the source of truth.
Monitors can live next to the infrastructure they watch, in the same pull request. Terraform, the CLI and the API all drive the same REST endpoints the dashboard uses, with write access on every plan. The dashboard is a view of that state, not where it lives.
Terraform
Provider on the Registry →resource "sentinel_monitor" "checkout_api" {
url = "https://api.acme.dev/health"
friendly_name = "Checkout API"
check_interval = 0.5
check_types = ["ssl", "json"]
}npm install -g @rootstuff/sentinel
sentinel monitors create \
--url https://api.acme.dev/health \
--interval 0.5curl -X POST \
https://sentinel.rootstuff.io/api/v1/monitors \
-H "Authorization: Bearer $SENTINEL_TOKEN" \
-H "Content-Type: application/json" \
-d '{"url": "https://api.acme.dev/health",
"check_interval": 0.5}'Incident context
Four questions, answered before you open your laptop.
When PagerDuty fires, the incident page should settle what happened, not start the investigation. Multi-region results, response times, deploy markers and the first failing check are already there.
Is it actually down?
The incident opened because at least eight of fifteen regions failed, spread across two continents or more. One region's bad route never gets this far.
The rule →
Is it regional?
Every region's last result sits on the monitor, with response times by region. A dark continent and a dead origin look different.
Per-region results →
Did something just deploy?
GitHub deployments are drawn on the charts and cited on the incident, so a deploy four minutes before the first failure is the first thing you see.
Deploy markers →
When did it start?
The incident records the first failing check, which regions saw it and what they saw: status code, timing phases, the error, and the response when it matters.
What's on an incident →
I already have Datadog or Grafana. What does Sentinel add?
A view from outside. APM and metrics tell you what the application sees once a request reaches it. They can't see a DNS record pointing at the old load balancer, a certificate that expired at the edge, a CDN serving its error page to one continent, or a WAF challenging your own users. Sentinel checks from fifteen regions on the public internet and alerts on what users get. It sits next to your observability stack, not in place of it, and pages the same PagerDuty service.
How does Sentinel avoid false alerts?
Every check runs from all fifteen regions. A failure in one region is recorded but doesn't open an incident on its own. Only when at least half of them, three or more on two continents, fail the same check does Sentinel mark the monitor down and send alerts, so a flaky route near one probe never alerts you.
What's the fastest check interval?
15 seconds on the Business plan. Pro checks every 30 seconds, Starter every minute, and the Free plan every five minutes. The interval is set per monitor, so you can run critical endpoints at the fastest cadence your plan allows and leave low-stakes endpoints slower.
Does it integrate with PagerDuty?
Yes. Sentinel opens a PagerDuty incident through the Events API v2 when a monitor goes down and resolves it on recovery. You connect a PagerDuty service to a team and choose which monitors route to it. PagerDuty is included on the Business plan.
Can I manage monitors with Terraform?
Yes. The Sentinel Terraform provider is on the Terraform Registry and covers monitors, groups and notification settings, so monitoring lives in the same pull requests as the infrastructure it watches. There's also a CLI on npm and a REST API with full read and write access on every plan.
Can it monitor APIs and things that aren't websites?
Yes. HTTP checks take any method, custom headers and body, basic, bearer or digest auth, the status codes you accept, and JSON assertions against the response. Beyond HTTP there are ping and TCP port monitors, DNS record and SSL certificate checks, and heartbeat monitors for cron jobs and queue workers that report in by pinging a URL.
Where do the checks come from?
From Sentinel's own probes in fifteen regions across six continents. The probe IPs are published so you can allow-list them, and the same probes are available on demand for a one-off check of any URL.
Alerts that mean something
Add your first endpoints in 30 seconds, or declare them in Terraform. Free to start, full API included.
No card for the free plan · 14-day trial on paid plans · Cancel anytime