X-Git-Url: http://git.squeep.com/?a=blobdiff_plain;f=lib%2Fpleroma%2Fweb%2Ftelemetry.ex;h=b0385060004cebed9b1083adec11e626d9feccea;hb=b98fe4476c13f213f5a4c928b7946516374e133b;hp=f7e9d5ca747fbebf34b559ebf7596f631ec4e722;hpb=33243c56e5560f99bd886ed44bcb50c9f679ccc2;p=akkoma

diff --git a/lib/pleroma/web/telemetry.ex b/lib/pleroma/web/telemetry.ex
index f7e9d5ca7..b03850600 100644
--- a/lib/pleroma/web/telemetry.ex
+++ b/lib/pleroma/web/telemetry.ex
@@ -1,6 +1,8 @@
 defmodule Pleroma.Web.Telemetry do
   use Supervisor
   import Telemetry.Metrics
+  alias Pleroma.Stats
+  alias Pleroma.Config
 
   def start_link(arg) do
     Supervisor.start_link(__MODULE__, arg, name: __MODULE__)
@@ -8,28 +10,31 @@ defmodule Pleroma.Web.Telemetry do
 
   @impl true
   def init(_arg) do
-    children = [
-      # Telemetry poller will execute the given period measurements
-      # every 10_000ms. Learn more here: https://hexdocs.pm/telemetry_metrics
-      {:telemetry_poller, measurements: periodic_measurements(), period: 10_000},
-      # Add reporters as children of your supervision tree.
-      # {Telemetry.Metrics.ConsoleReporter, metrics: metrics()},
-      {TelemetryMetricsPrometheus, metrics: metrics()}
-    ]
+    children =
+      [
+        {:telemetry_poller, measurements: periodic_measurements(), period: 10_000}
+      ] ++
+        prometheus_children()
 
     Supervisor.init(children, strategy: :one_for_one)
   end
 
-  def metrics do
+  defp prometheus_children do
+    config = Config.get([:instance, :export_prometheus_metrics], true)
+
+    if config do
+      [
+        {TelemetryMetricsPrometheus.Core, metrics: prometheus_metrics()},
+        Pleroma.PrometheusExporter
+      ]
+    else
+      []
+    end
+  end
+
+  # A seperate set of metrics for distributions because phoenix dashboard does NOT handle them well
+  defp distribution_metrics do
     [
-      # Phoenix Metrics
-      summary("phoenix.endpoint.stop.duration",
-        unit: {:native, :millisecond}
-      ),
-      summary("phoenix.router_dispatch.stop.duration",
-        tags: [:route],
-        unit: {:native, :millisecond}
-      ),
       distribution(
         "phoenix.router_dispatch.stop.duration",
         # event_name: [:pleroma, :repo, :query, :total_time],
@@ -37,7 +42,7 @@ defmodule Pleroma.Web.Telemetry do
         unit: {:native, :second},
         tags: [:route],
         reporter_options: [
-          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
+          buckets: [0.1, 0.2, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 250, 500, 1000]
         ]
       ),
 
@@ -48,7 +53,7 @@ defmodule Pleroma.Web.Telemetry do
         measurement: :total_time,
         unit: {:native, :millisecond},
         reporter_options: [
-          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
+          buckets: [0.1, 0.2, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 250, 500, 1000]
         ]
       ),
       distribution(
@@ -60,6 +65,52 @@ defmodule Pleroma.Web.Telemetry do
           buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
         ]
       ),
+      distribution(
+        "oban_job_exception",
+        event_name: [:oban, :job, :exception],
+        measurement: :duration,
+        tags: [:worker],
+        tag_values: fn tags -> Map.put(tags, :worker, tags.job.worker) end,
+        unit: {:native, :second},
+        reporter_options: [
+          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
+        ]
+      ),
+      distribution(
+        "tesla_request_completed",
+        event_name: [:tesla, :request, :stop],
+        measurement: :duration,
+        tags: [:response_code],
+        tag_values: fn tags -> Map.put(tags, :response_code, tags.env.status) end,
+        unit: {:native, :second},
+        reporter_options: [
+          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
+        ]
+      ),
+      distribution(
+        "oban_job_completion",
+        event_name: [:oban, :job, :stop],
+        measurement: :duration,
+        tags: [:worker],
+        tag_values: fn tags -> Map.put(tags, :worker, tags.job.worker) end,
+        unit: {:native, :second},
+        reporter_options: [
+          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
+        ]
+      )
+    ]
+  end
+
+  defp summary_metrics do
+    [
+      # Phoenix Metrics
+      summary("phoenix.endpoint.stop.duration",
+        unit: {:native, :millisecond}
+      ),
+      summary("phoenix.router_dispatch.stop.duration",
+        tags: [:route],
+        unit: {:native, :millisecond}
+      ),
       summary("pleroma.repo.query.total_time", unit: {:native, :millisecond}),
       summary("pleroma.repo.query.decode_time", unit: {:native, :millisecond}),
       summary("pleroma.repo.query.query_time", unit: {:native, :millisecond}),
@@ -71,20 +122,27 @@ defmodule Pleroma.Web.Telemetry do
       summary("vm.total_run_queue_lengths.total"),
       summary("vm.total_run_queue_lengths.cpu"),
       summary("vm.total_run_queue_lengths.io"),
-      distribution(
-        "oban.job.stop",
-        event_name: [:oban, :job, :stop],
-        measurement: :duration,
-        tags: [:job],
-        unit: {:native, :second},
-        reporter_options: [
-          buckets: [0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2.5, 5, 10]
-        ]
-      )
+      last_value("pleroma.local_users.total"),
+      last_value("pleroma.domains.total"),
+      last_value("pleroma.local_statuses.total"),
+      last_value("pleroma.remote_users.total")
     ]
   end
 
+  def prometheus_metrics, do: summary_metrics() ++ distribution_metrics()
+  def live_dashboard_metrics, do: summary_metrics()
+
   defp periodic_measurements do
-    []
+    [
+      {__MODULE__, :instance_stats, []}
+    ]
+  end
+
+  def instance_stats do
+    stats = Stats.get_stats()
+    :telemetry.execute([:pleroma, :local_users], %{total: stats.user_count}, %{})
+    :telemetry.execute([:pleroma, :domains], %{total: stats.domain_count}, %{})
+    :telemetry.execute([:pleroma, :local_statuses], %{total: stats.status_count}, %{})
+    :telemetry.execute([:pleroma, :remote_users], %{total: stats.remote_user_count}, %{})
   end
 end