From 7fce396e66096f8ccd41997e41b56219be2bce48 Mon Sep 17 00:00:00 2001
From: James Hu <hello@james.hu>
Date: Sun, 12 Nov 2023 16:45:13 -0800
Subject: [PATCH] Add optimizations and setting to retry failed Plex requests
 (#13)

---
 .github/workflows/docker.yml |  3 ++
 README.md                    |  2 ++
 docker-compose.yml           |  2 ++
 lib/middleware/collector.rb  | 66 ++++++++++++++++++++++++++++--------
 4 files changed, 59 insertions(+), 14 deletions(-)

diff --git a/.github/workflows/docker.yml b/.github/workflows/docker.yml
index 356e630..73ee663 100644
--- a/.github/workflows/docker.yml
+++ b/.github/workflows/docker.yml
@@ -53,8 +53,10 @@ jobs:
 
           echo 'EOF' >> $GITHUB_ENV
 
+      # Only build and push Docker images for releases
       - name: Build and push Docker image to GitHub Container Registry
         uses: docker/build-push-action@v2
+        if: ${{ github.event_name == 'release' }}
         with:
           context: .
           push: true
@@ -65,6 +67,7 @@ jobs:
           cache-to: type=local,dest=/tmp/.buildx-cache-new
 
       - name: Update Docker buildx cache
+        if: ${{ github.event_name == 'release' }}
         run: |
           rm -rf /tmp/.buildx-cache
           mv /tmp/.buildx-cache-new /tmp/.buildx-cache
diff --git a/README.md b/README.md
index 9421738..e3f0b39 100644
--- a/README.md
+++ b/README.md
@@ -19,6 +19,8 @@ These environment variables can be passed into the container (defaults are in pa
   - Plex Media Server token
 * `PLEX_TIMEOUT` (`10`)
   - How long to wait for Plex Media Server to respond
+* `PLEX_RETRIES_COUNT` (`0`)
+  - How many times to retry failed Plex Media Server requests
 * `METRICS_PREFIX` (`plex`)
   - What to prefix metric names with
 * `METRICS_MEDIA_COLLECTING_INTERVAL_SECONDS` (`300`)
diff --git a/docker-compose.yml b/docker-compose.yml
index 9fe604b..267c3f3 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -8,6 +8,8 @@ services:
       # Can be set in .env
       - PLEX_ADDR=${PLEX_ADDR}
       - PLEX_TOKEN=${PLEX_TOKEN}
+      - PLEX_TIMEOUT=${PLEX_TIMEOUT}
+      - PLEX_RETRIES_COUNT=${PLEX_RETRIES_COUNT}
       - METRICS_MEDIA_COLLECTING_INTERVAL_SECONDS=300
     volumes:
       - .:/srv:delegated
diff --git a/lib/middleware/collector.rb b/lib/middleware/collector.rb
index 3dd88aa..20689dc 100644
--- a/lib/middleware/collector.rb
+++ b/lib/middleware/collector.rb
@@ -9,7 +9,9 @@ def initialize(app)
 
         # Plex configs
         @plex_addr = ENV["PLEX_ADDR"] || "http://localhost:32400"
+        @plex_token = ENV["PLEX_TOKEN"]
         @plex_timeout = ENV["PLEX_TIMEOUT"]&.to_i || 10
+        @plex_retries_count = ENV["PLEX_RETRIES_COUNT"]&.to_i || 0
 
         # Metrics configs
         @metrics_prefix = ENV["METRICS_PREFIX"] || "plex"
@@ -54,21 +56,32 @@ def initialize(app)
       end
 
       def call(env)
+        case env["PATH_INFO"]
+        when "/metrics"
+          collect_metrics
+        end
+
+        @app.call(env)
+      end
+
+      def collect_metrics
+        log(step: "collect_metrics")
+
         begin
-          capabilities_resource = send_plex_api_request(method: :get, endpoint: "/").dig("MediaContainer")
+          capabilities_resource = send_plex_api_request(method: :get, endpoint: "/identity").dig("MediaContainer")
           metric_up_value = 1
 
+          log(plex_up: true)
+
           set_gauge_metric_values_or_reset_missing(
             metric: @metrics[:info],
             values: {
               { version: capabilities_resource.dig("version") } => 1,
             },
           )
-
-          collect_session_metrics
-          collect_activity_metrics
-          collect_media_metrics
         rescue HTTP::Error
+          log(plex_up: false)
+
           # Value of 0 means there's no heartbeat
           metric_up_value = 0
         ensure
@@ -80,11 +93,23 @@ def call(env)
           )
         end
 
-        @app.call(env)
+        [
+          -> { collect_session_metrics },
+          -> { collect_activity_metrics },
+          -> { collect_media_metrics },
+        ].each do |collection|
+          collection.call
+        rescue HTTP::Error
+          # Skip if it errors out
+        end
       end
 
       private
 
+      def log(**labels)
+        puts(labels.to_a.map { |k, v| "#{k}=#{v}" }.join(" "))
+      end
+
       def collect_activity_metrics
         values = Hash.new { |h, k| h[k] = 0 }
 
@@ -203,15 +228,28 @@ def fetch_media_section_count(key:, params: {}, **options)
       end
 
       def send_plex_api_request(method:, endpoint:, **options)
-        response = HTTP
-          .timeout(@plex_timeout)
-          .headers(
-            "X-Plex-Token" => ENV["PLEX_TOKEN"],
-            "Accept" => "application/json",
-          )
-          .public_send(method, "#{@plex_addr}#{endpoint}", **options)
+        retries_count = 0
 
-        JSON.parse(response)
+        # Keep trying request if it fails until number of retries have been exhausted
+        loop do
+          url = "#{@plex_addr}#{endpoint}"
+
+          log(method: method, url: url)
+
+          response = HTTP
+            .timeout(@plex_timeout)
+            .headers(
+              "X-Plex-Token" => @plex_token,
+              "Accept" => "application/json",
+            )
+            .public_send(method, url, **options)
+
+          return JSON.parse(response)
+        rescue HTTP::Error => e
+          raise(e) if retries_count >= @plex_retries_count
+
+          retries_count += 1
+        end
       end
 
       # Set metric values and reset all other labels that werenn't passed in