Instrument transient GitHub user refresh failures

app/models/shipit/user.rb

@@ -93,6 +93,13 @@ def stacks_contributed_to
       Commit.where('author_id = :id or committer_id = :id', id:).distinct.pluck(:stack_id)
     end
 
+    TRANSIENT_GITHUB_REFRESH_ERRORS = [
+      Faraday::ConnectionFailed,
+      Faraday::TimeoutError,
+      Net::OpenTimeout,
+      Net::ReadTimeout,
+    ].freeze
+
     def refresh_from_github!
       # Users are global, any app can be used
       # This will not work for users that only exist in an Enterprise install
@@ -101,6 +108,9 @@ def refresh_from_github!
       identify_renamed_user!
     rescue Octokit::Forbidden
       Rails.logger.info("User #{name}, github_id #{github_id} has forbidden access to their GitHub, likely deleted.")
+    rescue *TRANSIENT_GITHUB_REFRESH_ERRORS => error
+      instrument_transient_github_refresh_error(error)
+      raise
     end
 
     def github_user=(github_user)
@@ -155,6 +165,24 @@ def identify_renamed_user!
       false
     end
 
+    def instrument_transient_github_refresh_error(error)
+      payload = {
+        user_id: id,
+        github_id: github_id,
+        github_login: login,
+        error_class: error.class.name,
+        error_message: error.message,
+        operation: "refresh_github_user",
+      }
+
+      ActiveSupport::Notifications.instrument("transient_github_refresh_error.shipit", payload)
+      Rails.logger.warn(
+        "Transient GitHub user refresh error " \
+          "user_id=#{id} github_id=#{github_id} github_login=#{login.inspect} " \
+          "error_class=#{error.class.name} error_message=#{error.message.inspect}"
+      )
+    end
+
     def email_valid_and_preferred?(email_address)
       org_domains = Shipit.preferred_org_emails
       return true if org_domains.blank?

test/models/users_test.rb

@@ -212,6 +212,31 @@ class UsersTest < ActiveSupport::TestCase
       @user.refresh_from_github!
     end
 
+    test "#refresh_from_github! instruments transient GitHub failures and reraises" do
+      error = Net::OpenTimeout.new("execution expired")
+      events = []
+      subscriber = lambda do |*args|
+        events << ActiveSupport::Notifications::Event.new(*args)
+      end
+
+      Shipit.github.api.expects(:user).with(@user.github_id).raises(error)
+      Rails.logger.expects(:warn).with(regexp_matches(/Transient GitHub user refresh error/))
+
+      ActiveSupport::Notifications.subscribed(subscriber, "transient_github_refresh_error.shipit") do
+        assert_raises Net::OpenTimeout do
+          @user.refresh_from_github!
+        end
+      end
+
+      assert_equal 1, events.size
+      assert_equal @user.id, events.first.payload[:user_id]
+      assert_equal @user.github_id, events.first.payload[:github_id]
+      assert_equal @user.login, events.first.payload[:github_login]
+      assert_equal "Net::OpenTimeout", events.first.payload[:error_class]
+      assert_equal "execution expired", events.first.payload[:error_message]
+      assert_equal "refresh_github_user", events.first.payload[:operation]
+    end
+
     test "#github_api uses the user's access token" do
       assert_equal @user.github_access_token, @user.github_api.access_token
     end