Likely causeIncident 2291, opened 03:06

Deploy 4f2a1c raised the DB pool timeout.

87% confidence, explains 203 of 214 alerts

Logs

  1. deploy4f2a1c live on 12/12 pods
  2. configdb.pool.timeout 5s -> 30s
  3. appwaited 18.2s for a connection
  4. pgtoo many clients: 498 of 500

TracePOST /checkout2.41 s

  • auth.verify12 ms
  • cart.load38 ms
  • db.pool.acquire2.21 s
  • payments.charge140 ms

Daniel Reyes acknowledged at 03:12. Sam and Lena stayed asleep.

How it found this

p99 412 ms at 03:13

Arrow keys shove the pile. Enter runs the triage.

Six seconds between the storm and the page.

The same night, from 03:12:00. Pagerless does three things before anyone's phone lights up, and every one of them can be read back later.

  1. It reads every alert.

    Pagerless connects to the monitors you already run. Each alert becomes a service, a check, a value and a time, and repeats of the same alert become one line with a count.

    A laptop in a dark room, its screen full of scrolling log output
  2. It finds what changed.

    Alerts that moved together are lined up against the deploys, config changes and traces from the same minutes. The change that explains most of them becomes the likely cause, with a confidence you can check line by line.

    An aisle of server racks lit blue in a data centre
  3. It pages one person.

    The responder for that service gets one page with the cause card and the action that undoes it. Everyone else stays asleep. Nothing runs until a person presses the button.

    PAGERLESS

    checkout-api: likely cause found

    Deploy 4f2a1c raised the DB pool timeout. 87% confidence, roll back ready.

    • Daniel ReyesPrimaryPaged
    • Sam BeckerSecondaryAsleep
    • Lena WeissPaymentsAsleep

Works with what already pages you

Reads from Datadog, Prometheus, Grafana, CloudWatch, Sentry, Honeycomb and New Relic. Pages through its own app, PagerDuty, Opsgenie, Slack or Microsoft Teams. Rolls back with GitHub Actions, Argo CD, Spinnaker and kubectl.

Free for 5 responders.

Most teams start with one service and one rota, then add the rest once the first quiet night has happened.

Free

$0up to 5 responders

  • Every alert source
  • Cause cards with logs and traces
  • Roll back with approval
  • 30 days of incident history
Start free

Team

$19per responder a month

  • Everything in Free
  • Unlimited responders and rotas
  • Phone, SMS and push pages
  • 13 months of history and postmortem drafts
Start a 30-day trial

Scale

Customfrom 200 responders

  • Runs inside your own cloud account
  • SAML, SCIM and an audit log
  • Actions beyond roll back
  • A named engineer for your rollout
Book a call
Daniel Reyes at his laptop late at night

“I got one page instead of 214. It had the deploy, four log lines and the button. I was back asleep by 03:20.”

Daniel ReyesStaff SRE at Fernway

Questions from people who carry the pager.

Does Pagerless ever roll back on its own?

No. It proposes the action that undoes the likely cause and waits for a person. You can let it act alone on chosen services later, and every action is logged with the name of whoever approved it.

What does 87% confidence mean?

It is the share of the grouped alerts that the cause explains in time and place, weighted by how often the same pattern held in your own history. The card shows the log lines and the trace the score is built from, so you can disagree with it.

What happens when it names the wrong cause?

One tap pages the next person or opens the full alert list. The miss becomes a lesson for your workspace only, and the card says what it would check next.

Where does our data go?

Alert payloads and log lines are processed in your region and kept for 30 days on Free. On Scale, Pagerless runs inside your own cloud account and nothing leaves it.

How long does setup take?

Most teams connect one alert source and one rota in about 20 minutes. Pagerless then watches quietly for a week, shows what it would have paged, and only starts paging when you say so.