Remove OpenTelemetry for now

We're losing errors and it's possibly too much of an overhead for the moment in production environments.
nerves-hub · Dec 18, 2023 · dde177c · dde177c
1 parent 09bea32
commit dde177c
Show file tree

Hide file tree

Showing 11 changed files with 329 additions and 465 deletions.
diff --git a/config/config.exs b/config/config.exs
@@ -67,11 +67,6 @@ config :nerves_hub, NervesHubWeb.Endpoint,
   render_errors: [view: NervesHubWeb.ErrorView, accepts: ~w(html json)],
   pubsub_server: NervesHub.PubSub
 
-config :opentelemetry,
-  span_processor: :batch,
-  traces_exporter: :otlp,
-  resource: %{service: %{name: "nerves_hub"}}
-
 config :swoosh, :api_client, Swoosh.ApiClient.Finch
 
 # Environment specific config

diff --git a/config/dev.exs b/config/dev.exs
@@ -108,13 +108,4 @@ config :nerves_hub, NervesHubWeb.Endpoint,
     ]
   ]
 
-if System.get_env("OTEL_ENABLED", "false") == "true" do
-  # Export to a local collector
-  config :opentelemetry_exporter,
-    otlp_protocol: :http_protobuf,
-    otlp_endpoint: "http://localhost:4318"
-else
-  config :opentelemetry, tracer: :otel_tracer_noop, traces_exporter: :none
-end
-
 config :sentry, environment_name: :development
diff --git a/config/test.exs b/config/test.exs
@@ -83,12 +83,4 @@ config :nerves_hub, NervesHubWeb.Endpoint,
   secret_key_base: "x7Vj9rmmRke//ctlapsPNGHXCRTnArTPbfsv6qX4PChFT9ARiNR5Ua8zoRilNCmX",
   live_view: [signing_salt: "FnV9rP_c2BL11dvh"]
 
-# OTel
-config :opentelemetry, tracer: :otel_tracer_noop, traces_exporter: :none
-
-config :opentelemetry, :processors,
-  otel_batch_processor: %{
-    exporter: {:otel_exporter_tab, []}
-  }
-
 config :sentry, environment_name: :test
diff --git a/lib/nerves_hub/deployments.ex b/lib/nerves_hub/deployments.ex
@@ -2,7 +2,6 @@ defmodule NervesHub.Deployments do
   import Ecto.Query
 
   require Logger
-  require OpenTelemetry.Tracer, as: Tracer
 
   alias NervesHub.AuditLogs
   alias NervesHub.Deployments.Deployment
@@ -305,41 +304,33 @@ defmodule NervesHub.Deployments do
   Do nothing if a deployment is already set
   """
   def set_deployment(%{deployment_id: nil} = device) do
-    Tracer.with_span "Deployments.set_deployment" do
-      Tracer.set_attribute("nerves_hub.deployment.status", "setting")
-
-      case alternate_deployments(device, [true]) do
-        [] ->
-          Logger.debug("No matching deployments for #{device.identifier}")
-
-          %{device | deployment: nil}
-
-        [deployment] ->
-          device
-          |> Ecto.Changeset.change()
-          |> Ecto.Changeset.put_change(:deployment_id, deployment.id)
-          |> Repo.update!()
-          |> Repo.preload([:deployment])
-
-        [deployment | _] ->
-          Logger.debug(
-            "More than one deployment matches for #{device.identifier}, setting to the first"
-          )
-
-          device
-          |> Ecto.Changeset.change()
-          |> Ecto.Changeset.put_change(:deployment_id, deployment.id)
-          |> Repo.update!()
-          |> Repo.preload([:deployment])
-      end
+    case alternate_deployments(device, [true]) do
+      [] ->
+        Logger.debug("No matching deployments for #{device.identifier}")
+
+        %{device | deployment: nil}
+
+      [deployment] ->
+        device
+        |> Ecto.Changeset.change()
+        |> Ecto.Changeset.put_change(:deployment_id, deployment.id)
+        |> Repo.update!()
+        |> Repo.preload([:deployment])
+
+      [deployment | _] ->
+        Logger.debug(
+          "More than one deployment matches for #{device.identifier}, setting to the first"
+        )
+
+        device
+        |> Ecto.Changeset.change()
+        |> Ecto.Changeset.put_change(:deployment_id, deployment.id)
+        |> Repo.update!()
+        |> Repo.preload([:deployment])
     end
   end
 
   def set_deployment(device) do
-    Tracer.with_span "Deployments.set_deployment" do
-      Tracer.set_attribute("nerves_hub.deployment.status", "existed")
-
-      Repo.preload(device, [:deployment])
-    end
+    Repo.preload(device, [:deployment])
   end
 end
diff --git a/lib/nerves_hub/deployments/orchestrator.ex b/lib/nerves_hub/deployments/orchestrator.ex
@@ -11,7 +11,6 @@ defmodule NervesHub.Deployments.Orchestrator do
   use GenServer
 
   require Logger
-  require OpenTelemetry.Tracer, as: Tracer
 
   alias NervesHub.Devices
   alias NervesHub.Devices.Device
@@ -50,68 +49,57 @@ defmodule NervesHub.Deployments.Orchestrator do
   was successful, and the process is repeated.
   """
   def trigger_update(deployment) do
-    Tracer.with_span "Deploments.Orchestrator.trigger_update" do
-      :telemetry.execute([:nerves_hub, :deployment, :trigger_update], %{count: 1})
-
-      match_conditions = [
-        {:and, {:==, {:map_get, :deployment_id, :"$1"}, deployment.id},
-         {:==, {:map_get, :updating, :"$1"}, false},
-         {:==, {:map_get, :updates_enabled, :"$1"}, true},
-         {:"/=", {:map_get, :firmware_uuid, :"$1"}, deployment.firmware.uuid}}
-      ]
-
-      match_return = %{
-        device_id: {:element, 1, :"$_"},
-        pid: {:element, 1, {:element, 2, :"$_"}},
-        firmware_uuid: {:map_get, :firmware_uuid, {:element, 2, {:element, 2, :"$_"}}}
-      }
-
-      devices =
-        Registry.select(NervesHub.Devices, [
-          {{:_, :_, :"$1"}, match_conditions, [match_return]}
-        ])
-
-      # Get a rough count of devices to update
-      count = deployment.concurrent_updates - Devices.count_inflight_updates_for(deployment)
-      # Just in case inflight goes higher than concurrent, limit it to 0
-      count = max(count, 0)
-
-      Tracer.set_attributes(%{
-        "nerves_hub.deployments.devices_to_update" => count,
-        "nerves_hub.deployments.devices_online" => Enum.count(devices)
-      })
-
-      # use a reduce to bounce out early?
-      # limit the number of devices to 5 minutes / 500ms?
-
-      devices
-      |> Enum.take(count)
-      |> Enum.each(fn %{device_id: device_id, pid: pid} ->
-        Tracer.with_span "Deployments.Orchestrator.each_device" do
-          Tracer.set_attribute("nerves_hub.devices.id", device_id)
-
-          :telemetry.execute([:nerves_hub, :deployment, :trigger_update, :device], %{count: 1})
-
-          device = %Device{id: device_id}
-
-          # Check again because other nodes are processing at the same time
-          if Devices.count_inflight_updates_for(deployment) < deployment.concurrent_updates do
-            case Devices.told_to_update(device, deployment) do
-              {:ok, inflight_update} ->
-                send(pid, {"deployments/update", inflight_update})
-
-              :error ->
-                Logger.error(
-                  "An inflight update could not be created or found for the device #{device.identifier} (#{device.id})"
-                )
-            end
-          end
-
-          # Slow the update a bit to allow for concurrent nodes
-          Process.sleep(500)
+    :telemetry.execute([:nerves_hub, :deployment, :trigger_update], %{count: 1})
+
+    match_conditions = [
+      {:and, {:==, {:map_get, :deployment_id, :"$1"}, deployment.id},
+       {:==, {:map_get, :updating, :"$1"}, false},
+       {:==, {:map_get, :updates_enabled, :"$1"}, true},
+       {:"/=", {:map_get, :firmware_uuid, :"$1"}, deployment.firmware.uuid}}
+    ]
+
+    match_return = %{
+      device_id: {:element, 1, :"$_"},
+      pid: {:element, 1, {:element, 2, :"$_"}},
+      firmware_uuid: {:map_get, :firmware_uuid, {:element, 2, {:element, 2, :"$_"}}}
+    }
+
+    devices =
+      Registry.select(NervesHub.Devices, [
+        {{:_, :_, :"$1"}, match_conditions, [match_return]}
+      ])
+
+    # Get a rough count of devices to update
+    count = deployment.concurrent_updates - Devices.count_inflight_updates_for(deployment)
+    # Just in case inflight goes higher than concurrent, limit it to 0
+    count = max(count, 0)
+
+    # use a reduce to bounce out early?
+    # limit the number of devices to 5 minutes / 500ms?
+
+    devices
+    |> Enum.take(count)
+    |> Enum.each(fn %{device_id: device_id, pid: pid} ->
+      :telemetry.execute([:nerves_hub, :deployment, :trigger_update, :device], %{count: 1})
+
+      device = %Device{id: device_id}
+
+      # Check again because other nodes are processing at the same time
+      if Devices.count_inflight_updates_for(deployment) < deployment.concurrent_updates do
+        case Devices.told_to_update(device, deployment) do
+          {:ok, inflight_update} ->
+            send(pid, {"deployments/update", inflight_update})
+
+          :error ->
+            Logger.error(
+              "An inflight update could not be created or found for the device #{device.identifier} (#{device.id})"
+            )
         end
-      end)
-    end
+      end
+
+      # Slow the update a bit to allow for concurrent nodes
+      Process.sleep(500)
+    end)
   end
 
   def init(deployment) do

diff --git a/lib/nerves_hub/devices.ex b/lib/nerves_hub/devices.ex
@@ -23,8 +23,6 @@ defmodule NervesHub.Devices do
   alias NervesHub.Repo
   alias NervesHub.TaskSupervisor, as: Tasks
 
-  require OpenTelemetry.Tracer, as: Tracer
-
   @min_fwup_delta_updatable_version ">=1.6.0"
 
   def get_device(device_id), do: Repo.get(Device, device_id)
@@ -580,33 +578,21 @@ defmodule NervesHub.Devices do
   This may clear the deployment from the device if the version or tags are different.
   """
   def verify_deployment(%{deployment_id: nil} = device) do
-    Tracer.with_span "Devices.verify_deployment" do
-      Tracer.set_attribute("nerves_hub.device.deployment_id", nil)
-
-      device
-    end
+    device
   end
 
   def verify_deployment(device) do
-    Tracer.with_span "Devices.verify_deployment" do
-      Tracer.set_attribute("nerves_hub.device.deployment_id", device.deployment_id)
-
-      device = Repo.preload(device, [:deployment])
+    device = Repo.preload(device, [:deployment])
 
-      case matches_deployment?(device, device.deployment) do
-        true ->
-          Tracer.set_attribute("nerves_hub.device.match_deployment", true)
-
-          device
-
-        false ->
-          Tracer.set_attribute("nerves_hub.device.match_deployment", false)
+    case matches_deployment?(device, device.deployment) do
+      true ->
+        device
 
-          device
-          |> Ecto.Changeset.change()
-          |> Ecto.Changeset.put_change(:deployment_id, nil)
-          |> Repo.update!()
-      end
+      false ->
+        device
+        |> Ecto.Changeset.change()
+        |> Ecto.Changeset.put_change(:deployment_id, nil)
+        |> Repo.update!()
     end
   end