bitfab 0.64.9 → 0.64.11

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b270eed8dc307a13b2a5e98e723abb2a5433b0617eef2228c4a6c3e560cfc1c9
-  data.tar.gz: cb3d9ce6dfca20c7bfff8b36f9577cf23fda6a8eeaf1e4fdd811c3fd375a30b1
+  metadata.gz: ca0982d902b32b704685f3942b2aa7a4f1caee10db462d4560674a291e0139c2
+  data.tar.gz: afe8094259621f9719aa45c2446f682b5c3a485443625663f4ddfc5086b730da
 SHA512:
-  metadata.gz: 90d48fd92f1fd3eb0a913ea4b3070e677ac40fa0e5ec18b928fda31fa7e5e1158eaca47736d305eaee3a850800558cc2890d501596c3a37216ae8682f31acae7
-  data.tar.gz: 493ca207e1e61d5737f1ec8efb6126e8e2f0f118090f797ddf5d5d709349632ebd1c6de7673c67ec99142e5226b16e459f1c9e3addfa05504db72a1238e44868
+  metadata.gz: fc314e448b508c5ecef57cb62736e77faef740a830f56cd5bfd3cc29e5209dc97df84a8b7e94641f9cd4c23a34311db60d4e64695c87c77bc1078dc05a1df115
+  data.tar.gz: 13c217d4efd1704e20315748e4b710f3d7ab92e351c5cc5c6e13af98f2c008a8bbf65a7e9315b304618e5750c6ce748aadebdab1df26c4f2b9d2d7ba20106bb8

data/lib/bitfab/http_client.rb

@@ -83,6 +83,10 @@ module Bitfab
     end
 
     def get_simulation_plan
+      # Without a key the server answers 401, so skip the request entirely.
+      key = @api_key.respond_to?(:call) ? @api_key.call : @api_key
+      raise ArgumentError, SimulationPlan::NO_API_KEY if key.to_s.strip.empty?
+
       get("/api/sdk/sim-plan", timeout: SimulationPlan::READ_TIMEOUT)
     end
 

data/lib/bitfab/replay.rb

@@ -464,7 +464,8 @@ module Bitfab
       # trace-ID mapping complete_replay builds would otherwise race the
       # in-flight batches and hand back nil for every item.
       delivered_trace_ids = preserve_replay_failure(result_items, experiment_id, full_experiment_url) do
-        wait_for_replay_persistence(http_client, experiment_id, result_items.map { |item| item[:_sdk_trace_id] })
+        wait_for_replay_persistence(http_client, experiment_id, result_items.map { |item| item[:_sdk_trace_id] },
+          labels: replay_labels(result_items))
       end
 
       # Primary source for item[:trace_id]: the server's assigned traces.id, read
@@ -1243,7 +1244,7 @@ module Bitfab
     # repeat itself. Acks cannot settle a request that timed out client-side
     # after the server committed, so anything short of fully delivered falls
     # through to the server poll, which stays the authority there.
-    def wait_for_replay_persistence(http_client, experiment_id, sdk_trace_ids, timeout: nil)
+    def wait_for_replay_persistence(http_client, experiment_id, sdk_trace_ids, timeout: nil, labels: {})
       timeout ||= PERSISTENCE_TIMEOUT_SECONDS
       trace_ids = sdk_trace_ids.compact
 
@@ -1286,15 +1287,29 @@ module Bitfab
         sleep((deadline - Otel.monotonic_now).clamp(0, PERSISTENCE_POLL_SECONDS))
       end
 
-      cause = if flushed
-        ""
-      else
-        " Delivery was also not confirmed before the flush deadline, so the " \
-          "spans likely never reached the server."
+      missing.each do |trace_id|
+        warn_unconfirmed_replay(unconfirmed_replay_warning(labels.fetch(trace_id, trace_id), timeout, flushed))
+      end
+      read_back_trace_ids
+    end
+
+    def unconfirmed_replay_warning(label, seconds, flushed)
+      "[bitfab] replay #{label} could not confirm all spans reached Bitfab within #{format("%g", seconds)}s. " \
+        "The run continues; Bitfab will show this replay as capture incomplete." \
+        "#{" Some spans likely never left this process." unless flushed}"
+    end
+
+    def warn_unconfirmed_replay(message)
+      Bitfab.count_delivery_problem(message)
+      warn message
+    rescue Exception # rubocop:disable Lint/RescueException
+      nil
+    end
+
+    def replay_labels(items)
+      items.each_with_object({}) do |item, labels|
+        labels[item[:_sdk_trace_id]] = "#{item[:original_trace_id]}##{item[:attempt]}" if item[:_sdk_trace_id]
       end
-      raise "Replay traces were not fully persisted before the delivery deadline " \
-        "(experiment #{experiment_id}, missing #{missing.length} of " \
-        "#{expected_span_counts.length} trace(s)).#{cause}"
     end
 
     # Normalize a complete-replay tokens hash (string-keyed JSON) into the

data/lib/bitfab/replay_processes.rb

@@ -51,13 +51,13 @@ module Bitfab
           mock_overrides: overrides, dry_run: assignment.fetch("dry_run"))
         unless assignment["dry_run"]
           begin
-            mapped = Replay.wait_for_replay_persistence(http, assignment.fetch("experiment_id"), [result[:_sdk_trace_id]], timeout: delivery_timeout)
+            mapped = Replay.wait_for_replay_persistence(http, assignment.fetch("experiment_id"), [result[:_sdk_trace_id]],
+              timeout: delivery_timeout, labels: Replay.replay_labels([result]))
             result[:trace_id] = mapped[result[:_sdk_trace_id]] || result[:trace_id]
             if result[:_sdk_trace_id] && result[:trace_id].nil?
               status = http.get_replay_status(assignment.fetch("experiment_id"), {result[:_sdk_trace_id] => 1})
               result[:trace_id] = status.fetch("traceIds", {})[result[:_sdk_trace_id]]
             end
-            raise "Child replay trace persistence could not be confirmed" if result[:error].nil? && result[:trace_id].nil?
           rescue => error
             result[:replay_error] = error
             result[:error] = error.message

data/lib/bitfab/simulation_plan.rb

@@ -8,6 +8,7 @@ module Bitfab
     REFRESH_SECONDS = 60.0
     RETRY_SECONDS = 10.0
     MAX_HELD = 1000
+    NO_API_KEY = "no API key is set"
     @plans = ObjectSpace::WeakMap.new
     @plans_mutex = Mutex.new
     @plans_pid = Process.pid

data/lib/bitfab/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Bitfab
-  VERSION = "0.64.9"
+  VERSION = "0.64.11"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: bitfab
 version: !ruby/object:Gem::Version
-  version: 0.64.9
+  version: 0.64.11
 platform: ruby
 authors:
 - Harvest Team