DEV Community

Cover image for alive365: An Uptime Monitor That Checks Past the Login Page
uehara
uehara

Posted on

alive365: An Uptime Monitor That Checks Past the Login Page

A homepage can return 200 while nobody can log in. A plain uptime check will report that site as healthy.

alive365 is the uptime monitoring SaaS we build at EarthLink Network. It calls your sites and APIs on a schedule and runs a real browser through the login or checkout steps, so the "the page loads but nothing works" kind of outage shows up too. alive365 is not open for sign-up yet. This post shows what it does and who it is for, screen by screen, using our development environment; every target in it is a demo.

The short version

  • alive365 checks from the outside, on a schedule, that your site and API are up.
  • There are two kinds of checks: a browser check that replays a login or checkout in up to 50 steps, and an HTTP check that calls a URL (it can sign in first).
  • An outage becomes one incident, listed under "Incidents" in the dashboard. Repeated failures are added to it instead of opening new ones, and recovery is confirmed after five successes in a row. A daily report can go to email or a webhook (Slack, for example) every morning.
  • It fits small teams that run their own site, shop or SaaS with no one on monitoring duty. If you want to try it, contact us (end of section 9).

1. The monitor list and the two kinds of checks

The Monitors page lists every check: its type (HTTP or BROWSER), the URL, and how often it runs (Interval).

The alive365 Monitors page. At the top, "4/50 monitors" and a "New Monitor" button. Four monitors are listed: "Online shop - top page" (HTTP, https://shop.example.com/, 60s), "Orders API - health" (HTTP, https://api.example.com/health, 60s), "Company site" (HTTP, https://www.example.com/, 300s) and "Checkout - browser journey" (BROWSER, 900s). They were just created, so Uptime and Response show "—" and the status is "PENDING". Development environment, demo settings.

"New Monitor" asks for the type. Two are available:

  • HTTP Request calls a URL with GET or POST and checks that the status code is the one you expect, such as 200. Use it for API health endpoints and for pages that should open.
  • Browser Journey drives a real browser through a flow and checks that it reaches the end. More on this in section 2.

Intervals are 1, 3, 5, 10, 15, 30 or 60 minutes.

2. Browser checks for login and checkout

A 200 from the homepage does not tell you that the login button works or that the cart opens. A browser check does.

You list the steps a customer takes, and alive365 runs them in order in a real browser (Chromium, driven by Playwright). The steps are built in the Journey Builder.

The browser monitor wizard on "Journey Builder". Five steps are listed under "Journey Steps": 1 Navigate, URL https://shop.example.com/login; 2 Click, Selector a.sign-in; 3 Type, Selector #email, Value set to a demo email address; 4 Click, Selector button[type="submit"]; 5 Assert, Selector .account-menu, Condition visible. Below are buttons to add Navigate, Type, Click, Wait, Assert and Screenshot steps. Development environment, demo settings.

There are six step types:

  • Navigate opens a URL.
  • Type fills in a field.
  • Click presses a button or link.
  • Wait waits for a set time or until an element appears.
  • Assert checks that an element is on the page.
  • Screenshot captures the page at that point.

One journey can have up to 50 steps. If a step is missing its URL or selector, the screen shows "Journey is not valid" and will not move on. A password does not go into the step itself; the step can point to a stored secret with {{secrets.<name>}}.

The browser only starts when a check runs. Nothing sits running in between.

3. HTTP checks behind a login

Member pages and admin APIs need a login. The Authentication tab of an HTTP check signs in before the request.

The HTTP monitor wizard on "Request Configuration". The steps across the top are Basics, Request, Schedule, Checks, Alerts and Test, with Request selected. The request is GET https://members.example.com/account with "Follow redirects" on. Under the Headers, Authentication, Body and Advanced tabs, Authentication is open and "Form Login" is chosen from None, Basic Auth, Bearer Token, API Key and Form Login. Below the explanation are "Login URL" (https://members.example.com/login), "Username" and "Password". Development environment, demo settings.

The five options:

  • None: no login.
  • Basic Auth: a user name and password.
  • Bearer Token: a token, as most APIs use.
  • API Key: a key sent in a header you name, such as X-API-Key.
  • Form Login: posts the user name and password to the login form, keeps the session, then checks the target page.

Passwords and tokens are encrypted before they are stored, and neither the screen nor the API ever returns them afterwards.

"Login URL" can be left blank. alive365 then opens the target page without logging in and uses the page it is redirected to as the login page.

The encryption keys are separate for each environment and kept in AWS SSM Parameter Store. The stored value is decrypted only when a check runs.

Form Login is for ordinary login forms. As the screen says, sites built as single-page apps or protected by MFA cannot be logged into this way, and alive365 points you to a browser check instead.

4. What to check, and how often: an online shop

For a small online shop, a setup could look like this:

  • Homepage: HTTP check, every minute. The fastest way to learn the site itself is down.
  • Order API: HTTP check, every minute. Pages can load while order processing alone has stopped.
  • Member pages: HTTP check with Form Login, every 5 minutes, to confirm pages behind the login open.
  • Login to cart: browser check, every 15 minutes, to catch a button that does nothing or a step that never loads.

A browser check starts a real browser and works through the page step by step, so each run takes longer and costs more than an HTTP check. Leave the short intervals to HTTP checks, and use the browser check to confirm the whole flow still works at a slightly longer interval.

5. When something breaks: one incident per outage

Whether something is down is decided by fixed rules, not by AI. The same results always lead to the same decision. The defaults:

  • A failure opens one incident. The first failure marks it as a warning; two failures in a row raise it to error and mark the monitor DOWN.
  • While the incident is open, more failures are added to it instead of opening new ones. One outage does not turn into dozens of incidents.
  • Recovery is confirmed after five successes in a row, and the incident closes. That stops a flapping site from closing and reopening incidents over and over.

Incidents are listed under "Incidents" in the dashboard, each with a record of when it failed and when it recovered.

Here is how the defaults play out for a homepage checked every minute that goes down for 20 minutes:

  1. The first check after it breaks fails, and an incident opens as a warning.
  2. The next check fails too. The incident is raised to error and the monitor is marked DOWN.
  3. For the rest of the 20 minutes, each failure is added to the same incident.
  4. After the fix, the fifth success in a row confirms the recovery, and the incident closes.

The whole outage stays as a single incident. How long recovery takes to confirm depends on the interval: about 5 minutes at one-minute checks, about 50 minutes at ten-minute checks.

We keep AI out of this decision so the reason can always be explained the same way. "It failed twice in a row" is something you can tell a customer or a colleague as is. An AI judgment can give a different answer to the same results on a different day, and nobody can explain why.

One more rule: if a scheduled check does not run because of a problem on the alive365 side, it does not count as an outage of your site. It is recorded, but the monitor is not marked DOWN and no incident is opened. A hiccup in the monitoring service should not be recorded as your site going down.

6. Channels and the morning report

Channels live under Notifications. There are two kinds, EMAIL and WEBHOOK. For Slack, add a Slack incoming webhook URL as a webhook, and the report arrives in Slack.

Two channels on the alive365 "Notification Channels" page: "On-call email" (EMAIL, sent to a demo email address) and "Slack #ops" (WEBHOOK, url: https://hooks.slack.com/services…). Each has a Test button, an on/off switch and a "Daily report" switch. "Daily report" is on only for On-call email. Both show "0 monitors linked". Development environment, demo settings.

Turn on "Daily report" and that channel gets one message every morning at 9:00 a.m. Japan time, covering the last 24 hours for every monitor: successes, failures, uptime and current state. Messages to Slack are formatted for Slack.

The report tells you, in one message and without opening the dashboard, that every monitor ran and whether anything went down in the last 24 hours.

7. Try a change right away

Each monitor's page has a "Run now" button. It runs the check once, without waiting for the next scheduled run.

That test run is kept apart from the real record. It does not count toward uptime or the daily report, and it never opens an incident. After editing a browser journey, you can check it passes without touching your production numbers.

To avoid hammering the target, "Run now" on the same monitor will not start again until a short wait has passed since the last run.

8. Who it is for

alive365 fits teams like these:

  • You run your own site, shop or SaaS with a small team. Nobody is on monitoring duty, and no one checks every day that the site still opens.
  • A broken login or checkout costs you sales or support time. Knowing the homepage opens is not enough.
  • Nobody has time to open the dashboard every day. The morning report from section 6 tells you in one email or Slack message that every check ran and whether anything went down the day before.

alive365 checks from the outside. It is not a tool for watching CPU, memory or logs inside your servers. Even if you already have one of those, whether a customer can open the page and get through the flow still needs its own check, and that is the part alive365 covers.

9. What it runs on

alive365 runs on AWS. HTTP checks run on Lambda and SQS, browser checks run as ECS Fargate tasks that start only when a check runs, and records are kept in DynamoDB. With no servers left running, the cost stays small while the amount of monitoring is small.

Sign-in to the dashboard uses our own SSO, built on OAuth 2.0 with PKCE.

To try alive365, use our contact form, pick "Other inquiries" as the Inquiry Type, and tell us you would like to use alive365. It helps to mention how many sites or APIs you want to watch and which flows (login, checkout and so on) you want a browser to walk through.

Terms

  • Incident: one outage, from the moment it is detected until recovery is confirmed.
  • Webhook: a URL that receives a message's contents. A Slack incoming webhook is one example.
  • Status code: the three-digit number a web server returns. 200 means success; the 500s mean the server failed.
  • Selector: a way to point at a button or field on the page, such as #email or button[type="submit"].

Articles on alive365 — how it came to be, what it can do, and how it is built — will continue as a series. If you are interested, please like and follow.

alive365 is an uptime-monitoring SaaS that watches your websites and services. Learn more at https://alive365.site
Our other products are listed at https://www.eln.ne.jp/products


EarthLink Network builds its own products to run the whole company on AI — each one born from a need along the way. See the full index of what we build:

→ The 18 products EarthLink Network builds in-house

For more about the company and each product, visit www.eln.ne.jp.

Top comments (0)