rails_pod_kit 0.0.1 → 0.0.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: fb47aab3b005492f13eaaa6ede0a89a5891cfbc8d034ebc0f1284b590efbb59a
-  data.tar.gz: 3f0f50b521bb1f96655edcbb296ea00156f90b631239491c49fad381f351989e
+  metadata.gz: 24c0e428faf1a3c46646a1d7a0e693b5cde8373f6aea61f9dc005d6b6dcba192
+  data.tar.gz: 968fc08545c75f5b22f91c2c882068b20069d7bb2bb6f48d1abf2aa09f4bc71b
 SHA512:
-  metadata.gz: 33733cd886176b0341314e9ba6641cb0368f6010e3d24e0b11851d8a18d32119dd183eb8f17a885d8be5f4c5e36d7b123205d5051bf79b90e8f0cc95b434d0ef
-  data.tar.gz: db116a08eb1b3fb3c0d4017a5be0082568e48907032717d35dbb10933f741811b0009c64749937747aa44e45db0dbb73a1647d288f1dca986c6ec8007cc4531e
+  metadata.gz: 266b27abb0911f031e54522dfcea870d603e1ec143465b4441a206590654087c1604a9aeb4438745bc10600bf3696b7d4938c0c936e0145322b76cb0279a5513
+  data.tar.gz: dedc4665fb51d38b9e713caead650ee875108aed4a24e9f27bb3053f2ec0152bdcc0f3840cc0ac76056858f330b723c502bc5f66d726261db47a7033e6ae2142

data/README.md

@@ -8,7 +8,7 @@ packaged behind a single, opinionated entry point:
   worker (Sidekiq) processes — no sidecar, no separate collector process. A
   metrics agent (e.g. the Datadog Agent via OpenMetrics autodiscovery) scrapes
   the pod directly.
-- **Health checks** on `/healthz` (database, cache, Redis, optionally Sidekiq),
+- **Health checks** on `/healthz` (database, cache, optionally Redis and Sidekiq),
   wired for Kubernetes startup/liveness/readiness probes — a thin, opinionated
   wrapper around [health-monitor-rails](https://github.com/lbeder/health-monitor-rails).
 
@@ -90,8 +90,10 @@ no initializers.
 
 `Health.install!` configures health-monitor-rails with the kit's defaults:
 endpoint at `/healthz`, checking **database** (health_monitor's default),
-**cache**, **Redis** (connection injected by the host) and — when a `sidekiq:`
-thresholds hash is given — **Sidekiq**. The gem's Railtie mounts
+**cache**, — when a `redis:` connection is given — **Redis** (connection
+injected by the host) and — when a `sidekiq:` thresholds hash is given —
+**Sidekiq**. Omit `redis:` on hosts with no Redis dependency (e.g. a Solid
+Queue stack) to get a database + cache only endpoint. The gem's Railtie mounts
 `HealthMonitor::Engine` at `/` automatically; pass `mount: false` to keep
 route ownership (custom mount point, constraints) and mount it yourself in
 `config/routes.rb`. Any further tuning goes through the optional block, which
@@ -99,7 +101,9 @@ receives the `HealthMonitor` configuration.
 
 ```ruby
 RailsPodKit::Health.install!(
-  redis:   { url: ENV['REDIS_URL'] }, # or a ready Redis/ConnectionPool object
+  redis:   { url: ENV['REDIS_URL'] }, # optional: a ready Redis/ConnectionPool
+                                      # object or options hash; omit for no
+                                      # Redis provider
   path:    :healthz,                  # default
   sidekiq: { queue_size: 200, latency: 10.minutes }
 ) do |config|
@@ -107,6 +111,13 @@ RailsPodKit::Health.install!(
 end
 ```
 
+Without a `redis:` argument the endpoint checks only database and cache
+(plus Sidekiq when its thresholds are given):
+
+```ruby
+RailsPodKit::Health.install!(path: :healthz) # database + cache only
+```
+
 Probe wiring on Kubernetes:
 
 - **startup probe** → `/healthz` — the full check, so the pod doesn't go ready
@@ -321,7 +332,12 @@ Booting the full host app just to read a handful of Redis counters would cost
   require only the config core, so the Rails-free exporter stays
   railties-free and a `puma -C config/puma.rb` boot (which evaluates
   `config/puma.rb` before Rails) still gets the integrations once
-  `Bundler.require` runs.
+  `Bundler.require` runs. The main file also loads the yabeda-prometheus-mmap
+  adapter at require time — before any host initializer runs — so host
+  initializers may declare Yabeda metrics (yabeda-rails, yabeda-activejob, a
+  custom collector) safely: the adapter registers itself while no metric exists
+  yet, regardless of whether the app boots via `bin/rails server` or
+  `puma -C config/puma.rb`.
 
 ## Local verification
 

data/VERSION

@@ -1 +1 @@
-0.0.1
+0.0.3

data/lib/rails_pod_kit/health.rb

@@ -21,7 +21,7 @@ module RailsPodKit
   #   )
   #
   # gives the app a /healthz endpoint checking database (health_monitor's
-  # default), cache, Redis and (optionally) Sidekiq — suitable as a k8s startup
+  # default), cache and (optionally) Redis and Sidekiq — suitable as a k8s startup
   # probe. For liveness/readiness, probe `/healthz?providers[]=none` to
   # short-circuit the dependency checks once the pod is live, so a transient
   # Redis hiccup doesn't restart the app. Same trick for a manual basic check:
@@ -35,8 +35,11 @@ module RailsPodKit
 
     # Configures HealthMonitor with the kit's defaults.
     #
-    # redis:   Redis connection options hash (same shape the host passes to
-    #          Sidekiq) or a ready connection object (Redis / ConnectionPool).
+    # redis:   optional Redis connection options hash (same shape the host
+    #          passes to Sidekiq) or a ready connection object (Redis /
+    #          ConnectionPool); when given the Redis provider is added. Omit
+    #          entirely (or pass nil) on hosts with no Redis dependency to get
+    #          a database + cache only endpoint.
     # path:    mount-relative endpoint path (default :healthz).
     # sidekiq: optional thresholds hash; when given the Sidekiq provider is
     #          added with the provided `queue_size:` / `latency:` overrides
@@ -51,14 +54,17 @@ module RailsPodKit
     #
     # Any further host-specific tuning (extra providers, error callback, …) can
     # be done in the block, which receives the HealthMonitor configuration.
-    def install!(redis:, path: :healthz, sidekiq: nil, silence_controller_log: true, mount: true)
+    def install!(redis: nil, path: :healthz, sidekiq: nil, silence_controller_log: true, mount: true)
       @auto_mount = mount
 
       HealthMonitor.configure do |config|
         config.path = path
         config.cache
-        config.redis.configure do |redis_config|
-          redis_config.connection = build_connection(redis)
+
+        if redis
+          config.redis.configure do |redis_config|
+            redis_config.connection = build_connection(redis)
+          end
         end
 
         if sidekiq

data/lib/rails_pod_kit/puma.rb

@@ -29,8 +29,10 @@ module RailsPodKit
 
       # The Puma plugin only pulls in the exporter class; we must also load the
       # full mmap module so `Yabeda::Prometheus::Mmap.registry` exists and the
-      # mmap adapter is registered (it's a transitive dep, so Bundler.require in
-      # the host app doesn't auto-require it).
+      # mmap adapter is registered. The gem's main entry (lib/rails_pod_kit.rb)
+      # already loaded it at Bundler.require, so this is normally a no-op — it
+      # matters only under `puma -C config/puma.rb`, where config/puma.rb runs
+      # before Rails and thus before the main entry, so this is the first require.
       require 'yabeda/prometheus/mmap'
 
       # The control app exposes Puma's /stats over a localhost-only socket that

data/lib/rails_pod_kit.rb

@@ -43,6 +43,19 @@ require 'rails_pod_kit/config'
 #   - under `puma -C config/puma.rb` — where Puma evaluates config/puma.rb
 #     (and thus rails_pod_kit/puma) before Rails exists — this file is still
 #     fresh for Bundler.require, so the integrations load once Rails is up.
+
+# Load the mmap adapter here — at Bundler.require, before any host initializer
+# runs — so its self-registration (`Yabeda.register_adapter` at require time)
+# happens while no Yabeda metric exists yet, making it a no-op. The adapter's
+# own `mmap.rb` requires the adapter *before* defining
+# `Yabeda::Prometheus::Mmap.registry`; if a metric is already declared when the
+# adapter loads, registration eagerly reaches for that not-yet-defined method
+# and boots crash with a NoMethodError. Under `bin/rails server` Rails boots
+# (and its initializers may declare metrics) before config/puma.rb requires the
+# adapter, so loading it here is what makes the order-of-boot irrelevant. The
+# lazy require in RailsPodKit::Puma.activate then degrades to a cheap no-op.
+require 'yabeda/prometheus/mmap'
+
 require 'rails_pod_kit/sidekiq'
 require 'rails_pod_kit/health'
 require 'rails_pod_kit/railtie'

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: rails_pod_kit
 version: !ruby/object:Gem::Version
-  version: 0.0.1
+  version: 0.0.3
 platform: ruby
 authors:
 - Fabio Napoleoni