upright 0.2.0 → 0.3.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 1d1bb4a1f14f8e4dc16cf5bebcd5868917b2c0703f6dd94592d6f8c818dfa0b9
-  data.tar.gz: 4d3a0593d1f49f256145340f3e64c9ab6f467a759f1c0a60e81e58635528e0d1
+  metadata.gz: ccb561e794bdafa5e31caed2347f7d371d02e5230ca875c5f507d1d0620d7363
+  data.tar.gz: 3bc20100e850cdd93a81e58b7e9b6f3ef9377c44b17bbd7891dbe697699b23a6
 SHA512:
-  metadata.gz: 5d06f946c9874af94e42bca38bb58671e227b3c4f2ac2f9a523d36b3efca04771a3d0a087bb2644ea067e5a8094d35e1370ee85ded8ea140af084afbf4a4e29a
-  data.tar.gz: c2ac9d3d2822197887e22308dbdce7be1b47692825dcb25cbb660f0de8e24643788b2cd5a63c72c4863041b619a35e3b31b9e241c69861b59ff66c8398fcd9b2
+  metadata.gz: 474970b3ccbce05a396ceb2a3575a58272b602d679cf70d55d10538cc546aa31d6142e1de401ee5098c6c3a2733a624cba54e34b79a0dcd9c6af7ad7e8124f0d
+  data.tar.gz: 35ed0971379b1c50dd9583f95ed27aa2586f53002e8acd81c05851797d6f2d1c59ced8d6a08817613ecb89ab376d453a9697cf241b82be1409c11f1caa87c6d3

data/README.md

@@ -47,7 +47,8 @@ Upright is a self-hosted synthetic monitoring system. It provides a framework fo
 
 ## Installation
 
-Upright is designed to be run in it's own Rails app and deployed with Kamal.
+> [!NOTE]
+> Upright is designed to be run in its own Rails app and deployed with Kamal.
 
 ### Quick Start (New Project)
 
@@ -157,6 +158,91 @@ Upright.configure do |config|
 end
 ```
 
+### Probe Result Cleanup
+
+Upright automatically cleans up old probe results on a recurring schedule. You can configure the retention thresholds:
+
+```ruby
+Upright.configure do |config|
+  config.stale_success_threshold = 24.hours     # Delete successful results older than this (default: 24 hours)
+  config.stale_failure_threshold = 30.days       # Delete failed results older than this (default: 30 days)
+  config.failure_retention_limit = 20_000        # Keep at most this many failed results (default: 20,000)
+end
+```
+
+## Custom Probe Types
+
+Upright ships with four built-in probe types: HTTP, Playwright, SMTP, and Traceroute. You can register your own to extend the system.
+
+### 1. Register the type
+
+Add it to your initializer so Upright knows about its name and icon:
+
+```ruby
+# config/initializers/upright.rb
+Upright.configure do |config|
+  config.probe_types.register :ping, name: "Ping", icon: "📶"
+end
+```
+
+### 2. Create the probe class
+
+Add a Ruby class in your `probes/` directory that extends `FrozenRecord::Base` and includes `Upright::Probeable` and `Upright::ProbeYamlSource`. Implement `probe_type`, `probe_target`, `check`, and `on_check_recorded`:
+
+```ruby
+# probes/ping_probe.rb
+class PingProbe < FrozenRecord::Base
+  include Upright::Probeable
+  include Upright::ProbeYamlSource
+
+  stagger_by_site 3.seconds
+
+  def check
+    @ping_output, status = Open3.capture2e("ping", "-c", "1", "-W", "5", host)
+    status.success?
+  end
+
+  def on_check_recorded(probe_result)
+    if @ping_output.present?
+      Upright::Artifact.new(name: "ping.log", content: @ping_output).attach_to(probe_result)
+    end
+  end
+
+  def probe_type = "ping"
+  def probe_target = host
+end
+```
+
+The `check` method should return a truthy value on success and a falsy value on failure.
+
+### 3. Define probes in YAML
+
+Create a YAML file matching the class name (e.g., `PingProbe` → `probes/ping_probes.yml`):
+
+```yaml
+# probes/ping_probes.yml
+- name: "Cloudflare DNS"
+  host: "1.1.1.1"
+
+- name: "Google DNS"
+  host: "8.8.8.8"
+```
+
+Fields defined in the YAML are available as methods on the probe instance (e.g., `host`).
+
+### 4. Schedule it
+
+Add a recurring job in `config/recurring.yml`:
+
+```yaml
+production:
+  ping_probes:
+    command: "PingProbe.check_and_record_all_later"
+    schedule: every 30 seconds
+```
+
+The custom type will automatically appear in all dashboard dropdowns and filter links.
+
 ## Defining Probes
 
 ### HTTP Probes
@@ -171,12 +257,15 @@ Add probes to `probes/http_probes.yml`:
 - name: API Health
   url: https://api.example.com/health
   expected_status: 200
+  alert_severity: critical
 
 - name: Admin Panel
   url: https://admin.example.com
   basic_auth_credentials: admin_auth  # Key in Rails credentials
 ```
 
+The optional `alert_severity` field controls the Prometheus alert severity when a probe fails. Values: `medium`, `high` (default), `critical`.
+
 ### SMTP Probes
 
 Add probes to `probes/smtp_probes.yml`:
@@ -317,70 +406,82 @@ image: your-org/upright
 servers:
   web:
     hosts:
-      - nyc.upright.example.com
-    env:
-      tags:
-        SITE_SUBDOMAIN: nyc
-
-    hosts:
-      - ams.upright.example.com
-    env:
-      tags:
-        SITE_SUBDOMAIN: ams
-
+      - ams.upright.example.com: [amsterdam]
+      - nyc.upright.example.com: [new_york]
+      - sfo.upright.example.com: [san_francisco]
+  jobs:
     hosts:
-      - sfo.upright.example.com
-    env:
-      tags:
-        SITE_SUBDOMAIN: sfo
+      - ams.upright.example.com: [amsterdam]
+      - nyc.upright.example.com: [new_york]
+      - sfo.upright.example.com: [san_francisco]
+    cmd: bin/jobs
 
-registry:
-  server: ghcr.io
-  username: your-org
-  password:
-    - KAMAL_REGISTRY_PASSWORD
+proxy:
+  app_port: 3000
+  ssl: true
+  hosts:
+    - "*.upright.example.com"
 
 env:
-  clear:
-    RAILS_ENV: production
-    RAILS_LOG_TO_STDOUT: true
   secret:
     - RAILS_MASTER_KEY
-    - OIDC_CLIENT_ID
-    - OIDC_CLIENT_SECRET
+  tags:
+    amsterdam:
+      SITE_SUBDOMAIN: ams
+    new_york:
+      SITE_SUBDOMAIN: nyc
+    san_francisco:
+      SITE_SUBDOMAIN: sfo
 
 accessories:
   playwright:
-    image: mcr.microsoft.com/playwright:v1.55.0-noble
-    cmd: npx -y playwright run-server --port 53333
-    host: nyc.upright.example.com
-    port: 53333
-    env:
-      clear:
-        DEBUG: pw:api
+    image: jacoblincool/playwright:chromium-server-1.55.0
+    port: "127.0.0.1:53333:53333"
+    roles:
+      - jobs
+
+  prometheus:
+    image: prom/prometheus:v3.2.1
+    hosts:
+      - ams.upright.example.com
+    cmd: >-
+      --config.file=/etc/prometheus/prometheus.yml
+      --storage.tsdb.path=/prometheus
+      --storage.tsdb.retention.time=30d
+      --web.enable-otlp-receiver
+    files:
+      - config/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml
+      - config/prometheus/rules/upright.rules.yml:/etc/prometheus/rules/upright.rules.yml
+
+  alertmanager:
+    image: prom/alertmanager:v0.28.1
+    hosts:
+      - ams.upright.example.com
+    cmd: --config.file=/etc/alertmanager/alertmanager.yml
+    files:
+      - config/alertmanager/alertmanager.yml:/etc/alertmanager/alertmanager.yml
 ```
 
 ## Observability
 
 ### Prometheus
 
-The engine exposes metrics at `/metrics`. Configure Prometheus to scrape:
+Metrics are exposed via a Puma plugin at `http://0.0.0.0:9394/metrics`. Configure Prometheus to scrape:
 
 ```yaml
 scrape_configs:
   - job_name: upright
     static_configs:
-      - targets: ['localhost:3000']
-    metrics_path: /metrics
+      - targets: ['localhost:9394']
 ```
 
 ### Metrics Exposed
 
 - `upright_probe_duration_seconds` - Probe execution duration
-- `upright_probe_success` - Probe success/failure (1/0)
-- `upright_probe_status_code` - HTTP status code for HTTP probes
+- `upright_probe_up` - Probe status (1 = up, 0 = down)
+- `upright_http_response_status` - HTTP response status code
 
-Labels include: `probe_name`, `probe_type`, `site_code`, `site_city`, `site_country`
+Labels include: `type`, `name`, `site_code`, `site_city`, `site_country`
 
 ### AlertManager
 
@@ -391,20 +492,12 @@ groups:
   - name: upright
     rules:
       - alert: ProbeDown
-        expr: upright_probe_success == 0
-        for: 5m
-        labels:
-          severity: critical
-        annotations:
-          summary: "Probe {{ $labels.probe_name }} is down"
-
-      - alert: ProbeSlow
-        expr: upright_probe_duration_seconds > 10
+        expr: upright_probe_up == 0
         for: 5m
         labels:
-          severity: warning
+          severity: "{{ $labels.alert_severity }}"
         annotations:
-          summary: "Probe {{ $labels.probe_name }} is slow"
+          summary: "Probe {{ $labels.name }} is down"
 ```
 
 ### OpenTelemetry

data/app/assets/stylesheets/upright/uptime-bars.css

@@ -40,7 +40,6 @@
   /* Individual uptime bar */
   .uptime-bar {
     border-radius: 2px;
-    cursor: default;
     flex: 1;
     height: 24px;
     max-width: 12px;
@@ -48,6 +47,16 @@
     transition: transform 100ms, filter 100ms;
   }
 
+  a.uptime-bar {
+    cursor: pointer;
+    display: block;
+    text-decoration: none;
+  }
+
+  div.uptime-bar {
+    cursor: default;
+  }
+
   .uptime-bar:hover {
     filter: brightness(1.1);
     transform: scaleY(1.15);

data/app/controllers/upright/probe_results_controller.rb

@@ -12,6 +12,7 @@ class Upright::ProbeResultsController < Upright::ApplicationController
         .by_type(params[:probe_type])
         .by_status(params[:status])
         .by_name(params[:probe_name])
+        .by_date(params[:date])
         .with_attached_artifacts
     end
 end

data/app/controllers/upright/sessions_controller.rb

@@ -8,6 +8,7 @@ class Upright::SessionsController < Upright::ApplicationController
   end
 
   def create
+    reset_session
     user = Upright::User.from_omniauth(request.env["omniauth.auth"])
     session[:user_info] = { email: user.email, name: user.name }
     redirect_to upright.root_path

data/app/helpers/upright/probe_results_helper.rb

@@ -1,15 +1,7 @@
 module Upright::ProbeResultsHelper
-  PROBE_TYPE_ICONS = {
-    http: "🌐",
-    playwright: "🎭",
-    ping: "📶",
-    smtp: "✉️",
-    traceroute: "🛤️"
-  }
-
   def probe_type_icon(probe_type)
-    icon = PROBE_TYPE_ICONS.fetch(probe_type.to_s.downcase.to_sym)
-    content_tag(:span, icon, title: probe_type.titleize)
+    registered = Upright.probe_types.find(probe_type)
+    content_tag(:span, registered.icon, title: registered.name)
   end
 
   def type_filter_link(label, probe_type = nil)
@@ -44,6 +36,7 @@ module Upright::ProbeResultsHelper
     parts << "for #{params[:probe_type].titleize} probes" if params[:probe_type].present?
     parts << "named #{params[:probe_name]}" if params[:probe_name].present?
     parts << "with status #{params[:status]}" if params[:status].present?
+    parts << "on #{Date.parse(params[:date]).to_fs(:long)}" if params[:date].present?
     parts.join(" ")
   end
 end

data/app/models/concerns/upright/playwright/form_authentication.rb

@@ -19,9 +19,6 @@ module Upright::Playwright::FormAuthentication
     end
 
     def authenticator_for(service)
-      # First try the host app's authenticator, then fall back to engine's
       "::Playwright::Authenticator::#{service.to_s.camelize}".constantize
-    rescue NameError
-      "Upright::Playwright::Authenticator::#{service.to_s.camelize}".constantize
     end
 end

data/app/models/concerns/upright/probe_result/stale_cleanup.rb

@@ -0,0 +1,23 @@
+module Upright::ProbeResult::StaleCleanup
+  extend ActiveSupport::Concern
+
+  class_methods do
+    def cleanup_stale
+      cleanup_stale_successes
+      cleanup_stale_failures
+    end
+
+    def cleanup_stale_successes
+      ok.where(created_at: ...Upright.config.stale_success_threshold.ago).in_batches.destroy_all
+    end
+
+    def cleanup_stale_failures
+      cutoff = [
+        Upright.config.stale_failure_threshold.ago,
+        fail.order(created_at: :desc).offset(Upright.config.failure_retention_limit).pick(:created_at)
+      ].compact.max
+
+      fail.where(created_at: ..cutoff).in_batches.destroy_all
+    end
+  end
+end

data/app/models/concerns/upright/probeable.rb

@@ -2,7 +2,7 @@ module Upright::Probeable
   extend ActiveSupport::Concern
   include Upright::Staggerable
 
-  TYPES = %w[ http playwright smtp traceroute ]
+  ALERT_SEVERITIES = %i[ medium high critical ]
 
   included do
     attr_writer :logger
@@ -35,6 +35,7 @@ module Upright::Probeable
       probe_name: probe_name,
       probe_target: probe_target,
       probe_service: probe_service,
+      probe_alert_severity: probe_alert_severity,
       status: result[:status],
       duration: result[:duration],
       error: result[:error]
@@ -63,6 +64,11 @@ module Upright::Probeable
     nil
   end
 
+  def probe_alert_severity
+    severity = try(:alert_severity)&.to_sym
+    ALERT_SEVERITIES.include?(severity) ? severity : :high
+  end
+
   private
     def failsafe_check
       result, error, duration = nil

data/app/models/upright/http/request.rb

@@ -23,7 +23,7 @@ class Upright::HTTP::Request
         proxy: proxy_url,
         proxyuserpwd: proxy_userpwd,
         verbose: true,
-        forbid_reuse: true
+        forbid_reuse: proxy_url.nil?
       }.compact
     end
 

data/app/models/upright/probe_result.rb

@@ -1,11 +1,16 @@
 class Upright::ProbeResult < Upright::ApplicationRecord
   include Upright::ExceptionRecording
+  include Upright::ProbeResult::StaleCleanup
+
+  attr_accessor :probe_alert_severity
 
   has_many_attached :artifacts
 
   scope :by_type,   ->(type) { where(probe_type: type) if type.present? }
   scope :by_status, ->(status) { where(status: status) if status.present? }
   scope :by_name,   ->(name) { where(probe_name: name) if name.present? }
+  scope :by_date,   ->(date) { where(created_at: Date.parse(date).all_day) if date.present? }
+
   scope :stale,     -> { where(created_at: ...24.hours.ago) }
 
   enum :status, [ :ok, :fail ]
@@ -23,7 +28,7 @@ class Upright::ProbeResult < Upright::ApplicationRecord
 
   private
     def increment_metrics
-      labels = { type: probe_type, name: probe_name, probe_target: probe_target, probe_service: probe_service }
+      labels = { type: probe_type, name: probe_name, probe_target: probe_target, probe_service: probe_service, alert_severity: probe_alert_severity || :high }
 
       Yabeda.upright_probe_duration_seconds.set(labels.merge(status: status), duration.to_f)
       Yabeda.upright_probe_up.set(labels, ok? ? 1 : 0)

data/app/models/upright/probes/status.rb

@@ -17,9 +17,9 @@ class Upright::Probes::Status
       end
 
       def label_selector(probe_type)
-        if probe_type.present?
-          "{type=\"#{probe_type}\"}"
-        end
+        matchers = [ "alert_severity!=\"\"" ]
+        matchers << "type=\"#{probe_type}\"" if probe_type.present?
+        "{#{matchers.join(",")}}"
       end
 
       def prometheus_client

data/app/models/upright/probes/uptime.rb

@@ -17,13 +17,13 @@ class Upright::Probes::Uptime
 
     private
       def query(probe_type)
-        "upright:probe_uptime_daily#{label_selector(probe_type)}"
+        "min by (name, type, probe_target) (upright:probe_uptime_daily#{label_selector(probe_type)})"
       end
 
       def label_selector(probe_type)
-        if probe_type.present?
-          "{type=\"#{probe_type}\"}"
-        end
+        matchers = [ "alert_severity!=\"\"" ]
+        matchers << "type=\"#{probe_type}\"" if probe_type.present?
+        "{#{matchers.join(",")}}"
       end
 
       def prometheus_client

data/app/views/upright/dashboards/_uptime_probe_row.html.erb

@@ -1,5 +1,5 @@
 <div class="data-table__row uptime-bars__row">
-  <%= link_to upright.site_root_url(subdomain: Upright.sites.first.code, probe_type: probe.type, probe_name: probe.name), class: "data-table__probe" do %>
+  <%= link_to upright.site_root_url(subdomain: current_or_default_site.code, probe_type: probe.type, probe_name: probe.name), class: "data-table__probe" do %>
     <%= probe_type_icon(probe.type) %>
     <span class="data-table__probe-name"><%= probe.name %></span>
   <% end %>
@@ -11,8 +11,10 @@
       <% else %>
         <% uptime_percent = uptime * 100 %>
         <% downtime_minutes = ((1 - uptime) * 24 * 60).round %>
-        <div class="uptime-bar uptime-bar--<%= uptime_label(uptime_percent) %>"
-             title="<%= uptime_bar_tooltip(date, uptime_percent, downtime_minutes) %>"></div>
+        <%= link_to "", upright.site_root_url(subdomain: current_or_default_site.code, probe_type: probe.type, probe_name: probe.name, date: date.iso8601),
+              class: "uptime-bar uptime-bar--#{uptime_label(uptime_percent)}",
+              title: uptime_bar_tooltip(date, uptime_percent, downtime_minutes),
+              aria: { label: uptime_bar_tooltip(date, uptime_percent, downtime_minutes) } %>
       <% end %>
     <% end %>
   </div>

data/app/views/upright/dashboards/probe_statuses/show.html.erb

@@ -4,7 +4,7 @@
 
     <div class="dashboard-filters">
       <%= form_with url: dashboards_probe_status_path, method: :get, data: { controller: "form", turbo_frame: :probe_status_matrix } do %>
-        <%= collection_select nil, :probe_type, Upright::Probeable::TYPES, :itself, :titleize,
+        <%= collection_select nil, :probe_type, Upright.probe_types, :type, :name,
               { selected: @probe_type },
               { class: "input--select", data: { action: "change->form#submit" } } %>
       <% end %>

data/app/views/upright/dashboards/uptimes/show.html.erb

@@ -4,7 +4,7 @@
 
     <div class="dashboard-filters">
       <%= form_with url: dashboards_uptime_path, method: :get, data: { controller: "form", turbo_frame: :uptime_bars } do %>
-        <%= collection_select nil, :probe_type, Upright::Probeable::TYPES, :itself, :titleize,
+        <%= collection_select nil, :probe_type, Upright.probe_types, :type, :name,
               { selected: @probe_type },
               { class: "input--select", data: { action: "change->form#submit" } } %>
       <% end %>

data/app/views/upright/probe_results/index.html.erb

@@ -2,10 +2,9 @@
   <aside>
     <ul>
       <li><%= type_filter_link "All" %></li>
-      <li><%= type_filter_link "HTTP", "http" %></li>
-      <li><%= type_filter_link "Playwright", "playwright" %></li>
-      <li><%= type_filter_link "SMTP", "smtp" %></li>
-      <li><%= type_filter_link "Traceroute", "traceroute" %></li>
+      <% Upright.probe_types.each do |probe_type| %>
+        <li><%= type_filter_link probe_type.name, probe_type.type %></li>
+      <% end %>
     </ul>
   </aside>
 
@@ -16,6 +15,7 @@
     <div class="filters">
       <%= form_with url: site_root_path, method: :get, data: { controller: "form" } do |f| %>
         <%= hidden_field_tag :probe_type, params[:probe_type] if params[:probe_type].present? %>
+        <%= hidden_field_tag :date, params[:date] if params[:date].present? %>
         <label for="probe_name">Name</label>
         <%= select_tag :probe_name,
               options_for_select([["All", ""]] + @probe_names, params[:probe_name]),
@@ -26,6 +26,9 @@
               options_for_select([["All", ""]] + Upright::ProbeResult.statuses.keys.map { |s| [s.titleize, s] }, params[:status]),
               class: "input--select",
               data: { action: "change->form#submit" } %>
+        <% if params[:date].present? %>
+          <%= link_to "Clear date filter", site_root_path(probe_type: params[:probe_type], probe_name: params[:probe_name], status: params[:status]), class: "filter-clear" %>
+        <% end %>
       <% end %>
     </div>
 

data/config/ci.rb

@@ -3,5 +3,7 @@
 CI.run do
   step "Style: Ruby", "bin/rubocop"
   step "Security: Brakeman code analysis", "bin/brakeman --quiet --no-pager --exit-on-warn --exit-on-error"
+  step "Lint: GitHub Actions (actionlint)", "actionlint"
+  step "Lint: GitHub Actions (zizmor)", "zizmor ."
   step "Tests: Rails", "bin/rails test"
 end

data/lib/generators/upright/install/install_generator.rb

@@ -5,6 +5,15 @@ module Upright
 
       desc "Install Upright engine into your application"
 
+      def set_ruby_version
+        ruby_version = RUBY_VERSION
+        create_file ".ruby-version", "#{ruby_version}\n"
+        create_file "mise.toml", <<~TOML
+          [tools]
+          ruby = "#{ruby_version}"
+        TOML
+      end
+
       def copy_initializers
         template "upright.rb", "config/initializers/upright.rb"
         template "omniauth.rb", "config/initializers/omniauth.rb"
@@ -43,6 +52,38 @@ module Upright
         template "puma.rb", "config/puma.rb"
       end
 
+      def install_solid_queue
+        rails_command "solid_queue:install"
+      end
+
+      def add_queue_database
+        gsub_file "config/database.yml",
+          "development:\n  <<: *default\n  database: storage/development.sqlite3",
+          "development:\n  primary:\n    <<: *default\n    database: storage/development.sqlite3\n  queue:\n    <<: *default\n    database: storage/development_queue.sqlite3\n    migrations_paths: db/queue_migrate"
+
+        gsub_file "config/database.yml",
+          "test:\n  <<: *default\n  database: storage/test.sqlite3",
+          "test:\n  primary:\n    <<: *default\n    database: storage/test.sqlite3\n  queue:\n    <<: *default\n    database: storage/test_queue.sqlite3\n    migrations_paths: db/queue_migrate"
+      end
+
+      def copy_recurring_config
+        template "recurring.yml", "config/recurring.yml", force: true
+      end
+
+      def add_jobs_to_procfile
+        procfile = File.join(destination_root, "Procfile.dev")
+        if File.exist?(procfile)
+          unless File.read(procfile).include?("jobs:")
+            append_to_file "Procfile.dev", "jobs: OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES bin/rails solid_queue:start\n"
+          end
+        else
+          create_file "Procfile.dev", <<~PROCFILE
+            web: bin/rails server -b '0.0.0.0' -p ${PORT:-3000}
+            jobs: OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES bin/rails solid_queue:start
+          PROCFILE
+        end
+      end
+
       def add_routes
         route 'mount Upright::Engine => "/", as: :upright'
       end
@@ -60,14 +101,14 @@ module Upright
         say "Upright has been installed!", :green
         say ""
         say "Next steps:"
-        say "  1. Run migrations: bin/rails db:migrate"
+        say "  1. Prepare the database: bin/rails db:prepare"
         say "  2. Configure your servers in config/deploy.yml"
         say "  3. Configure sites in config/sites.yml"
         say "  4. Add probes in probes/*.yml"
         say "  5. Set ADMIN_PASSWORD env var (default: upright)"
         say ""
         say "For production, review config/initializers/upright.rb and update:"
-        say "  config.hostname = \"honcho-upright.com\""
+        say "  config.hostname = \"example.com\""
         say ""
         say "Start dev services (Prometheus, Alertmanager, Playwright):"
         say "  docker compose up -d"

data/lib/generators/upright/install/templates/http_probes.yml

@@ -7,6 +7,7 @@
 #
 # Optional fields:
 #   - expected_status: HTTP status code to expect (default: 200-399)
+#   - alert_severity: Alert severity when probe fails: medium, high (default), critical
 #   - basic_auth_credentials: Key in Rails credentials for basic auth
 #   - proxy: Key in Rails credentials for proxy settings
 

data/lib/generators/upright/install/templates/recurring.yml

@@ -0,0 +1,25 @@
+development:
+  http_probes:
+    schedule: every minute
+    command: "Upright::Probes::HTTPProbe.check_and_record_all_later"
+
+  smtp_probes:
+    schedule: every minute
+    command: "Upright::Probes::SMTPProbe.check_and_record_all_later"
+
+production:
+  http_probes:
+    schedule: "*/30 * * * * *"
+    command: "Upright::Probes::HTTPProbe.check_and_record_all_later"
+
+  smtp_probes:
+    schedule: "*/30 * * * * *"
+    command: "Upright::Probes::SMTPProbe.check_and_record_all_later"
+
+  clear_solid_queue_finished_jobs:
+    command: "SolidQueue::Job.clear_finished_in_batches(sleep_between_batches: 0.3)"
+    schedule: every hour at minute 12
+
+  cleanup_stale_probe_results:
+    command: "Upright::ProbeResult.cleanup_stale"
+    schedule: every hour at minute 30

data/lib/generators/upright/install/templates/smtp_probes.yml

@@ -4,6 +4,9 @@
 # Required fields:
 #   - name: Unique identifier for the probe
 #   - host: SMTP server hostname
+#
+# Optional fields:
+#   - alert_severity: Alert severity when probe fails: medium, high (default), critical
 
 # - name: Example Mail Server
 #   host: mail.example.com

data/lib/generators/upright/install/templates/traceroute_probes.yml

@@ -4,6 +4,9 @@
 # Required fields:
 #   - name: Unique identifier for the probe
 #   - host: Hostname or IP address to trace
+#
+# Optional fields:
+#   - alert_severity: Alert severity when probe fails: medium, high (default), critical
 
 # - name: Example Service
 #   host: example.com

data/lib/generators/upright/install/templates/upright.rb

@@ -18,4 +18,7 @@ Upright.configure do |config|
   #   client_id: ENV["OIDC_CLIENT_ID"],
   #   client_secret: ENV["OIDC_CLIENT_SECRET"]
   # }
+
+  # Register custom probe types (built-in types: http, playwright, smtp, traceroute)
+  # config.probe_types.register :ftp_file, name: "FTP File", icon: "📂"
 end

data/lib/generators/upright/install/templates/upright.rules.yml

@@ -8,9 +8,9 @@ groups:
       # Fraction of regions reporting DOWN (0.0 to 1.0)
       - record: upright:probe_down_fraction
         expr: |
-          count by (name, type, probe_target) (upright_probe_up == 0)
+          count by (name, type, probe_target, alert_severity) (upright_probe_up == 0)
           /
-          count by (name, type, probe_target) (upright_probe_up)
+          count by (name, type, probe_target, alert_severity) (upright_probe_up)
 
       # Daily uptime percentage (0.0 to 1.0)
       # Uptime = percentage of time in past day when majority of sites reported UP
@@ -148,7 +148,7 @@ groups:
           upright: "https://app.<%= app_domain %>/?probe_type={{ $labels.type }}&status=fail&probe_name={{ $labels.name | urlquery }}"
         expr: upright:probe_down_fraction{type="http"} > 0.5
         labels:
-          severity: page
+          severity: "{{ $labels.alert_severity }}"
           group: upright
 
       - alert: UprightHTTPProbeDegraded
@@ -187,7 +187,7 @@ groups:
           upright: "https://app.<%= app_domain %>/?probe_type={{ $labels.type }}&status=fail&probe_name={{ $labels.name | urlquery }}"
         expr: upright:probe_down_fraction{type="smtp"} > 0.5
         labels:
-          severity: page
+          severity: "{{ $labels.alert_severity }}"
           group: upright
 
       - alert: UprightSMTPProbeDegraded
@@ -226,7 +226,7 @@ groups:
           upright: "https://app.<%= app_domain %>/?probe_type={{ $labels.type }}&status=fail&probe_name={{ $labels.name | urlquery }}"
         expr: upright:probe_down_fraction{type="playwright"} > 0.5
         labels:
-          severity: page
+          severity: "{{ $labels.alert_severity }}"
           group: upright
 
       - alert: UprightPlaywrightProbeDegraded

data/lib/upright/configuration.rb

@@ -29,6 +29,14 @@ class Upright::Configuration
   attr_accessor :prometheus_url
   attr_accessor :alert_webhook_url
 
+  # Probe types
+  attr_reader :probe_types
+
+  # Probe result cleanup
+  attr_accessor :stale_success_threshold
+  attr_accessor :stale_failure_threshold
+  attr_accessor :failure_retention_limit
+
   def initialize
     @service_name = "upright"
     @user_agent = "Upright/1.0"
@@ -41,11 +49,17 @@ class Upright::Configuration
     @probes_path = nil
     @authenticators_path = nil
 
+    @probe_types = Upright::ProbeTypeRegistry.new
+
     @playwright_server_url = ENV["PLAYWRIGHT_SERVER_URL"]
     @otel_endpoint = ENV["OTEL_EXPORTER_OTLP_ENDPOINT"]
 
     @auth_provider = :static_credentials
     @auth_options = {}
+
+    @stale_success_threshold = 24.hours
+    @stale_failure_threshold = 30.days
+    @failure_retention_limit = 20_000
   end
 
   def global_subdomain

data/lib/upright/engine.rb

@@ -56,6 +56,13 @@ class Upright::Engine < ::Rails::Engine
     end
   end
 
+  initializer "upright.probe_types", before: :load_config_initializers do
+    Upright.config.probe_types.register :http, name: "HTTP", icon: "🌐"
+    Upright.config.probe_types.register :playwright, name: "Playwright", icon: "🎭"
+    Upright.config.probe_types.register :smtp, name: "SMTP", icon: "✉️"
+    Upright.config.probe_types.register :traceroute, name: "Traceroute", icon: "🛤️"
+  end
+
   initializer "upright.frozen_record" do
     FrozenRecord::Base.base_path = Upright.configuration.frozen_record_path
   end

data/lib/upright/metrics.rb

@@ -44,12 +44,12 @@ module Upright::Metrics
             gauge :probe_duration_seconds,
               comment: "Duration of each probe",
               aggregation: :max,
-              tags: %i[type name probe_target probe_service status]
+              tags: %i[type name probe_target probe_service alert_severity status]
 
             gauge :probe_up,
               comment: "Probe status (1 = up, 0 = down)",
               aggregation: :most_recent,
-              tags: %i[type name probe_target probe_service]
+              tags: %i[type name probe_target probe_service alert_severity]
 
             gauge :http_response_status,
               comment: "HTTP response status code",

data/lib/upright/probe_type_registry.rb

@@ -0,0 +1,33 @@
+module Upright
+  class ProbeTypeRegistry
+    ProbeType = Data.define(:type, :name, :icon)
+
+    include Enumerable
+
+    def initialize
+      @probe_types = []
+    end
+
+    def register(type, name:, icon:)
+      type = type.to_s
+      @probe_types.reject! { |pt| pt.type == type }
+      @probe_types << ProbeType.new(type:, name:, icon:)
+    end
+
+    def types
+      @probe_types.map(&:type)
+    end
+
+    def find(type)
+      @probe_types.find { |pt| pt.type == type.to_s }
+    end
+
+    def unregister(type)
+      @probe_types.reject! { |pt| pt.type == type.to_s }
+    end
+
+    def each(&block)
+      @probe_types.each(&block)
+    end
+  end
+end

data/lib/upright/version.rb

@@ -1,3 +1,3 @@
 module Upright
-  VERSION = "0.2.0"
+  VERSION = "0.3.0"
 end

data/lib/upright.rb

@@ -19,6 +19,7 @@ require "yabeda/puma/plugin"
 
 require "upright/version"
 require "upright/configuration"
+require "upright/probe_type_registry"
 require "upright/geohash"
 require "upright/site"
 require "upright/metrics"
@@ -38,6 +39,10 @@ module Upright
       yield(configuration)
     end
 
+    def probe_types
+      configuration.probe_types
+    end
+
     def sites
       @sites ||= load_sites
     end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: upright
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.3.0
 platform: ruby
 authors:
 - Lewis Buckley
@@ -404,6 +404,7 @@ files:
 - app/models/concerns/upright/playwright/logging.rb
 - app/models/concerns/upright/playwright/otel_tracing.rb
 - app/models/concerns/upright/playwright/video_recording.rb
+- app/models/concerns/upright/probe_result/stale_cleanup.rb
 - app/models/concerns/upright/probe_yaml_source.rb
 - app/models/concerns/upright/probeable.rb
 - app/models/concerns/upright/staggerable.rb
@@ -464,6 +465,7 @@ files:
 - lib/generators/upright/install/templates/otel_collector.yml
 - lib/generators/upright/install/templates/prometheus.yml
 - lib/generators/upright/install/templates/puma.rb
+- lib/generators/upright/install/templates/recurring.yml
 - lib/generators/upright/install/templates/sites.yml
 - lib/generators/upright/install/templates/smtp_probes.yml
 - lib/generators/upright/install/templates/traceroute_probes.yml
@@ -480,6 +482,7 @@ files:
 - lib/upright/geohash.rb
 - lib/upright/metrics.rb
 - lib/upright/playwright/collect_performance_metrics.js
+- lib/upright/probe_type_registry.rb
 - lib/upright/site.rb
 - lib/upright/tracing.rb
 - lib/upright/version.rb
@@ -504,7 +507,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 4.0.3
+rubygems_version: 4.0.6
 specification_version: 4
 summary: Synthetic monitoring engine with Playwright and Prometheus metrics
 test_files: []