Collector fails to send payload over HTTP2 with the {:exceeds_window_size, :request, 65536} error

[arkgil]

Hi there 👋

At the end of the test suite, the collector fails with the following stacktrace:

** (CaseClauseError) no case clause matching: {:error, %Mint.HTTP2{buffer: <<0, 0, 4, 8, 0, 0, 0, 0, 0, 127, 255, 0, 0>>, client_settings: %{enable_push: true, max_concurrent_streams: 100, max_frame_size: 16384}, client_settings_queue: {[[]], []}, decode_table: %HPAX.Table{entries: [], length: 0, max_table_size: 4096, size: 0}, encode_table: %HPAX.Table{entries: [], length: 0, max_table_size: 4096, size: 0}, headers_being_processed: nil, hostname: "analytics-api.buildkite.com", mode: :active, next_stream_id: 3, open_client_stream_count: 0, open_server_stream_count: 0, ping_queue: {[], []}, port: 443, private: %{}, proxy_headers: [], ref_to_stream_id: %{}, scheme: "https", server_settings: %{enable_connect_protocol: false, enable_push: true, initial_window_size: 65536, max_concurrent_streams: 128, max_frame_size: 16777215, max_header_list_size: :infinity}, socket: {:sslsocket, {:gen_tcp, #Port<0.335>, :tls_connection, :undefined}, [#PID<0.3828.0>, #PID<0.3827.0>]}, state: :open, streams: %{}, transport: Mint.Core.Transport.SSL, window_size: 65535}, %Mint.HTTPError{module: Mint.HTTP2, reason: {:exceeds_window_size, :request, 65536}}}
--
  | (buildkite_test_collector 0.1.0) lib/buildkite_test_collector/http_transport.ex:18: BuildkiteTestCollector.HttpTransport.send/1
  | (buildkite_test_collector 0.1.0) lib/buildkite_test_collector/formatter.ex:104: BuildkiteTestCollector.Formatter.terminate/2
  | (stdlib 3.17) gen_server.erl:733: :gen_server.try_terminate/3
  | (stdlib 3.17) gen_server.erl:918: :gen_server.terminate/10
  | (stdlib 3.17) proc_lib.erl:226: :proc_lib.init_p_do_apply/3

This looks like the error described in elixir-tesla/tesla#394.

[jimsynz]

Hi @arkgil thanks for the report.

I'll take a look at fixing this first thing in the morning. The payloads are already batched (in groups of 500) so worst case we can reduce the batch size to fix.

[arkgil]

Can I somehow configure the batch size to try if it helps?

[jimsynz]

Yup. You can add :batch_size to your :buildkite_test_collector configuration in test.exs eg:

import Config

config :buildkite_test_collector,
  api_key: "my redacted key",
  batch_size: 100

Hope that helps!

[arkgil]

Thanks! Will try it out.

[arkgil]

Alright, batch size of 100 fixes the issues 👍

[joshprice]

Closing this for now, we'll reduce default batch size to prevent this from occurring, and investigate ways to resolve the underlying problem too. Thanks for raising @arkgil ❤️

[ryansch]

I just hit this on 0.1.2 as well. I'll try lowering the batch size.

[joshprice]

Thanks for reporting. I've updated the default to 100 (34901e4), but please let us know what value(s) worked for you.

This will be in the next release.

[arkgil]

@joshprice yup, we've set the batch size to 100, not issues so far. Thanks!