hutch 1.3.1 → 2.0.0.rc1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 02b687d56ec0d56dc49c07722e3e55c3b86ca274008394209b3fe60d6ecb8bf4
-  data.tar.gz: a6c38c9b5f26944f7abb3b078d4cf693bb29ed2fb08987b7f4bc5e4fe2a70f78
+  metadata.gz: 8978faf08cf85e68cc957d91ff3059b0b3fecb73e018a80ec551c9547be9c99a
+  data.tar.gz: 2c1f461463321e705e2464373093135bf307b68dac9f08b284e8c5f6c9e710db
 SHA512:
-  metadata.gz: 85513bc753214f4a077c4a50510519082e3f5bc8aee75cba1df09461be9723e4a8d438a97f3c6bd3025fb59a0d7008adb5577c7faee082367ddd85f7239f748a
-  data.tar.gz: f0d6fcc77c94a0349104243c3daac3e2f4fe057ce96a4b251c0577969fcaf1b69860ca3eadd163aac353275bbe517847573f6064ea46cad70aa8979004c6e9f7
+  metadata.gz: 8900ac9233d8849f789b2f598e8109d8bcc7aab3bc7870c00d881e47aaa6f3a3722b906a018c8c8f6dd0b69258a15f7f7654027a216429a9bfe2639f53070100
+  data.tar.gz: e3c8b5ac42f24483a773167eb79d01e9e29e71859c0a78ccf01d59b018ec1f57a1a5c26c8da47ad9ebb743b04fed9d67989802eb0f495bd894c1b3c51cb830c3

data/CHANGELOG.md

@@ -1,8 +1,112 @@
-## 1.3.2 (in development)
+# Hutch Change Log
 
-No changes yet.
+## 2.0.0 (TBD)
 
-## 1.3.1 (Dec 11, 2024)
+### Breaking Changes
+
+ * Ruby `3.2` or JRuby `10` is now required
+ * The Datadog tracer now depends on the [`datadog`](https://github.com/DataDog/dd-trace-rb)
+   gem, which replaces the EOL `ddtrace`
+   ([migration guide](https://github.com/DataDog/dd-trace-rb/blob/master/docs/UpgradeGuide2.md))
+ * `Hutch::ErrorHandlers::SentryRaven` has been replaced by
+   `Hutch::ErrorHandlers::Sentry`, backed by `sentry-ruby`
+   ([migration guide](https://docs.sentry.io/platforms/ruby/migration/))
+ * Messages are serialized with the stdlib `json` instead of `multi_json`, which
+   means that `MultiJson` configuration such as `MultiJson.use(:oj)` no longer
+   affects Hutch
+
+### Consumer Recovery After a RabbitMQ 4.3 Quorum Queue Consumer Timeout
+
+RabbitMQ 4.3 quorum queues cancel a timed out consumer with `basic.cancel`
+instead of closing the channel. Hutch 1.4.0 does not react to that, so on 4.3
+a consumer stops consuming its queue until the worker restarts.
+
+Workers now re-subscribe consumers cancelled by the server, but only if
+the queue still exists: a `basic.cancel` carries no reason, so a consumer
+timeout is indistinguishable from a queue deletion. The consumer channel is
+replaced in the process, so the deliveries the cancelled consumer still held
+are requeued instead of dropped.
+
+Handlers that are still running get a graceful exit to finish before
+the old channel is closed, and their acknowledgements are dropped rather than
+sent on the new channel, where their delivery tags mean nothing.
+
+### Connection URIs Are Now Percent-Decoded
+
+The virtual host, username and password of a connection URI are now decoded, as
+[the URI specification](https://www.rabbitmq.com/docs/uri-spec) requires.
+`amqp://host/%2F` connects to the default virtual host instead of one literally
+named `%2F`, and credentials containing a space, `#` or `@` authenticate again.
+
+Reported by @deissimon.
+
+GitHub issue: [#422](https://github.com/ruby-amqp/hutch/issues/422)
+
+### Fixed Two JRuby Regressions
+
+Opening a channel raised, and the `march_hare` adapter did not declare the
+exchange it instantiated.
+
+GitHub issue: [#427](https://github.com/ruby-amqp/hutch/issues/427)
+
+### The JRuby Variant Is Revived
+
+The `java` platform gem, which depends on March Hare instead of Bunny, was last
+published as `0.25.0` in January 2018 and has seen little more than CI version
+bumps since. This release requires JRuby `10` and `march_hare` `4.7.0`, adds a
+JRuby CI job, and publishes the gem again.
+
+Recovering a channel the broker closed remains Bunny-only: `MarchHare::Channel`
+has no `on_error` callback and no `reopen`.
+
+## 1.4.0 (Apr 7, 2026)
+
+### Ruby 3.0 is Now Required
+
+This version adopts [Bunny `3.x`](https://github.com/ruby-amqp/bunny/releases/tag/3.0.0)
+and as a result, requires Ruby 3.0.
+
+### `without_namespace` Consumer DSL Option
+
+Consumers can now opt out of the automatic namespace prefix
+for their queue name using the `without_namespace` DSL method:
+
+``` ruby
+class DeadLetterConsumer
+  include Hutch::Consumer
+  consume 'dead.letters'
+  queue_name 'deadletter'
+  without_namespace
+end
+```
+
+Originally contributed by @tlloydthwaites.
+
+GitHub issue: [#393](https://github.com/ruby-amqp/hutch/pull/393)
+
+### Consumer Channel Recovery After Delivery Acknowledgement Timeout
+
+Hutch now automatically "recovers" (reopens) consumer channels closed by RabbitMQ
+due to a [delivery acknowledgement timeout](https://www.rabbitmq.com/docs/consumers#acknowledgement-timeout).
+
+Note that this exception indicates a potential issue on Hutch's end:
+a delivery was not acknowledged in time (30 minutes by default). Even then,
+recovering the channel in this specific scenario is an operational improvement.
+
+Contributed by @Garaio-REM.
+
+GitHub issue: [#414](https://github.com/ruby-amqp/hutch/pull/414)
+
+### Migrated Datadog Tracer From the `ddtrace` Gem to `datadog`
+
+The Datadog tracer now uses the `datadog` gem instead of the deprecated `ddtrace`.
+The `ddtrace` gem is still supported but emits a deprecation warning at load time.
+Support for `ddtrace` will be removed in Hutch 2.0.
+
+### Deprecated `SentryRaven` Error Handler
+
+`Hutch::ErrorHandlers::SentryRaven` now emits a deprecation warning and will be
+removed in Hutch 2.0. Use `Hutch::ErrorHandlers::Sentry` (backed by `sentry-ruby`) instead.
 
 ### Rails 8.x Compatibility
 
@@ -19,7 +123,7 @@ GitHub issue: [#392](https://github.com/ruby-amqp/hutch/pull/392)
 
 ### Relaxed ActiveSupport Dependency Constraints
 
-Contributed by @drobny.
+Contributed by drobny.
 
 GitHub issue: [#402](https://github.com/ruby-amqp/hutch/pull/402)
 
@@ -93,7 +197,7 @@ contains potentially breaking changes.
 
    This means **some defaults introduced in `0.28.0` ([gocardless/hutch#341](https://github.com/gocardless/hutch/pull/341)) were reverted**.
    The user has to opt in to configure the queue type and mode and other [optional arguments](https://www.rabbitmq.com/queues.html#optional-arguments) they need to use.
-   Most optional arguments can be set via [policies](https://www.rabbitmq.com/parameters.html#policies) which is always the recommended approach. 
+   Most optional arguments can be set via [policies](https://www.rabbitmq.com/parameters.html#policies) which is always the recommended approach.
    Queue type, unfortunately, is not one of them as different queue types have completely different
    implementation details, on disk data formats and so on.
 
@@ -106,7 +210,7 @@ contains potentially breaking changes.
       # when in doubt, prefer using a policy to this DSL
       # https://www.rabbitmq.com/parameters.html#policies
       arguments 'x-key': :value
-        
+
       quorum_queue
    end
    ```
@@ -120,7 +224,7 @@ contains potentially breaking changes.
      # when in doubt, prefer using a policy to this DSL
       # https://www.rabbitmq.com/parameters.html#policies
       arguments 'x-key': :value
-     
+
      lazy_queue
      classic_queue
    end

data/LICENSE

@@ -1,5 +1,5 @@
 Copyright (c) 2013-2016 GoCardless
-Copyright (c) 2016-2021 Hutch contributors
+Copyright (c) 2016-2026 Hutch contributors
 
 Permission is hereby granted, free of charge, to any person
 obtaining a copy of this software and associated documentation

data/README.md

@@ -1,7 +1,5 @@
-![](http://cl.ly/image/3h0q3F3G142K/hutch.png)
-
-[![Gem Version](https://badge.fury.io/rb/hutch.svg)](http://badge.fury.io/rb/hutch)
-[![Code Climate](https://codeclimate.com/github/gocardless/hutch.svg)](https://codeclimate.com/github/gocardless/hutch)
+[![Gem Version](https://badge.fury.io/rb/hutch.svg)](https://badge.fury.io/rb/hutch)
+[![CI](https://github.com/ruby-amqp/hutch/actions/workflows/test.yml/badge.svg)](https://github.com/ruby-amqp/hutch/actions/workflows/test.yml)
 
 Hutch is a Ruby library for enabling asynchronous inter-service communication
 in a service-oriented architecture, using RabbitMQ.
@@ -23,6 +21,8 @@ gem install hutch
     - [Message Processing Tracers](#message-processing-tracers)
   - [Running Hutch](#running-hutch)
     - [Loading Consumers](#loading-consumers)
+    - [Consumer Groups](#consumer-groups)
+    - [Loading Consumers Manually (One-by-One)](#loading-consumers-manually-one-by-one)
     - [Stopping Hutch](#stopping-hutch)
   - [Producers](#producers)
     - [Producer Configuration](#producer-configuration)
@@ -38,8 +38,8 @@ gem install hutch
 
 ## Requirements
 
-- Hutch requires Ruby 2.4+ or JRuby 9K.
-- Hutch requires RabbitMQ 3.3 or later.
+- Hutch requires Ruby 3.2+ or JRuby 10+
+- Hutch requires RabbitMQ 3.13 or later
 
 ## Overview
 
@@ -52,7 +52,7 @@ With Hutch, consumers are stored in separate files and include the `Hutch::Consu
 They are then loaded by a command line runner which connects to RabbitMQ, sets up queues and bindings,
 and so on. Publishers connect to RabbitMQ via `Hutch.connect` and publish using `Hutch.publish`.
 
-Hutch uses [Bunny](http://rubybunny.info) or [March Hare](http://rubymarchhare.info)
+Hutch uses [Bunny](https://github.com/ruby-amqp/bunny) or [March Hare](https://github.com/ruby-amqp/march_hare)
 (on JRuby) under the hood.
 
 ### Project Maturity
@@ -79,7 +79,7 @@ message[:id]  # => "02ABCXYZ"
 To subscribe to a topic, pass a routing key to `consume` in the class
 definition. To bind to multiple routing keys, simply pass extra routing keys
 in as additional arguments. Refer to the [RabbitMQ docs on topic exchanges
-](http://www.rabbitmq.com/tutorials/tutorial-five-ruby.html) for more information
+](https://www.rabbitmq.com/tutorials/tutorial-five-ruby) for more information
 about how to use routing keys. Here's an example consumer:
 
 ```ruby
@@ -110,10 +110,10 @@ end
 
 It is possible to set some custom options to consumer's queue explicitly.
 This example sets the consumer's queue as a
-[quorum queue](https://www.rabbitmq.com/quorum-queues.html)
-and to operate in the [lazy mode](https://www.rabbitmq.com/lazy-queues.html).
+[quorum queue](https://www.rabbitmq.com/docs/quorum-queues)
+and to operate in the [lazy mode](https://www.rabbitmq.com/docs/lazy-queues).
 The `initial_group_size`
-[argument](https://www.rabbitmq.com/quorum-queues.html#replication-factor) is
+[argument](https://www.rabbitmq.com/docs/quorum-queues#replication-factor) is
 optional.
 
 ```ruby
@@ -142,11 +142,11 @@ end
 ```
 
 This sets the `x-max-length` header. For more details, see the [RabbitMQ
-documentation on Queue Length Limit](https://www.rabbitmq.com/maxlength.html). To find out more
-about custom queue arguments, consult the [RabbitMQ documentation on AMQP Protocol Extensions](https://www.rabbitmq.com/extensions.html).
+documentation on Queue Length Limit](https://www.rabbitmq.com/docs/maxlength). To find out more
+about custom queue arguments, consult the [RabbitMQ documentation on AMQP Protocol Extensions](https://www.rabbitmq.com/docs/extensions).
 
 Consumers can write to Hutch's log by calling the logger method. The logger method returns
-a [Logger object](http://ruby-doc.org/stdlib-2.1.2/libdoc/logger/rdoc/Logger.html).
+a [Logger object](https://docs.ruby-lang.org/en/3.0/Logger.html).
 
 ```ruby
 class FailedPaymentConsumer
@@ -174,7 +174,7 @@ client_logger = Logger.new("/path/to/bunny.log")
 Hutch::Config.set(:client_logger, client_logger)
 ```
 
-See this [RabbitMQ tutorial on topic exchanges](http://www.rabbitmq.com/tutorials/tutorial-five-ruby.html)
+See this [RabbitMQ tutorial on topic exchanges](https://www.rabbitmq.com/tutorials/tutorial-five-ruby)
 to learn more.
 
 ### Message Processing Tracers
@@ -206,7 +206,7 @@ usage: hutch [options]
         --mq-host HOST               Set the RabbitMQ host
         --mq-port PORT               Set the RabbitMQ port
     -t, --[no-]mq-tls                Use TLS for the AMQP connection
-        --mq-tls-cert FILE           Certificate  for TLS client verification
+        --mq-tls-cert FILE           Certificate for TLS client verification
         --mq-tls-key FILE            Private key for TLS client verification
         --mq-exchange EXCHANGE       Set the RabbitMQ exchange
         --mq-vhost VHOST             Set the RabbitMQ vhost
@@ -242,13 +242,13 @@ in the config file, allowing for easy customization.
 
 Using Hutch with a Rails app is simple. Either start Hutch in the working
 directory of a Rails app, or pass the path to a Rails app in with the
-`--require` option. Consumers defined in Rails apps should be placed with in
+`--require` option. Consumers defined in Rails apps should be placed within
 the `app/consumers/` directory, to allow them to be auto-loaded when Rails
 boots.
 
-If you're using the new Zeitwerk autoloader (enabled by default in Rails 6)
-and the consumers are not loaded in development environment you will need to
-trigger the autoloading in an initializer with
+If you're using the Zeitwerk autoloader (the default since Rails 6)
+and the consumers are not loaded in the development environment, you will need to
+trigger eager loading in an initializer with
 
 ```ruby
 ::Zeitwerk::Loader.eager_load_all
@@ -309,8 +309,6 @@ kill -SIGTERM 456 # or kill -15 456
 kill -SIGQUIT 789 # or kill -3 789
 ```
 
-![](http://g.recordit.co/wyCdzG9Kh3.gif)
-
 ## Producers
 
 Hutch includes a `publish` method for sending messages to Hutch consumers. When
@@ -334,7 +332,7 @@ Hutch::Config.set(:mq_exchange, 'name')
 ### Publisher Confirms
 
 For maximum message reliability when producing messages, you can force Hutch to use
-[Publisher Confirms](https://www.rabbitmq.com/confirms.html) and wait for a confirmation
+[Publisher Confirms](https://www.rabbitmq.com/docs/confirms) and wait for a confirmation
 after every message published. This is the safest possible option for publishers
 but also results in a **significant throughput drop**.
 
@@ -354,25 +352,27 @@ send messages to Hutch.
 - Hutch works with topic exchanges, check the producer is also using topic
   exchanges.
 - Use message routing keys that match those used in your Hutch consumers.
-- Be sure your exchanges are marked as durable. In the Ruby AMQP gem, this is
-  done by passing `durable: true` to the exchange creation method.
-- Publish messages as persistent.
-- Using publisher confirms is highly recommended.
+- Be sure your exchanges are marked as durable, by passing `durable: true`
+  to the exchange creation method
+- Publish messages as persistent
+- Using publisher confirms is highly recommended
 
 Here's an example of a well-behaved publisher, minus publisher confirms:
 
 ```ruby
-AMQP.connect(host: config[:host]) do |connection|
-  channel  = AMQP::Channel.new(connection)
-  exchange = channel.topic(config[:exchange], durable: true)
+require 'bunny'
 
-  message = JSON.dump({ subject: 'Test', id: 'abc' })
-  exchange.publish(message, routing_key: 'test', persistent: true)
-end
-```
+conn = Bunny.new(host: config[:host])
+conn.start
 
-If using publisher confirms with amqp gem, see [this issue](https://github.com/ruby-amqp/amqp/issues/92)
-and [this gist](https://gist.github.com/3042381) for more info.
+ch       = conn.create_channel
+exchange = ch.topic(config[:exchange], durable: true)
+
+message = JSON.dump({ subject: 'Test', id: 'abc' })
+exchange.publish(message, routing_key: 'test', persistent: true)
+
+conn.close
+```
 
 ## Configuration
 
@@ -385,7 +385,7 @@ Known configuration parameters are:
  * `mq_host`: RabbitMQ hostname (default: `localhost`)
  * `mq_port`: RabbitMQ port (default: `5672`)
  * `mq_vhost`: vhost to use (default: `/`)
- * `mq_username`: username to use (default: `guest`, only can connect from localhost as of RabbitMQ 3.3.0)
+ * `mq_username`: username to use (default: `guest`, can only connect from localhost)
  * `mq_password`: password to use (default: `guest`)
  * `mq_tls`: should TLS be used? (default: `false`)
  * `mq_tls_cert`: path to client TLS certificate (public key)
@@ -407,7 +407,7 @@ Known configuration parameters are:
     exception happens, see classes in `Hutch::Acknowledgements`.
  * `mq_exchange`: exchange to use for publishing (default: `hutch`)
  * `mq_client_properties`: Bunny's [client properties](https://www.rabbitmq.com/docs/connections#capabilities) (default: `{}`)
- * `heartbeat`: [RabbitMQ heartbeat timeout](http://rabbitmq.com/heartbeats.html) (default: `30`)
+ * `heartbeat`: [RabbitMQ heartbeat timeout](https://www.rabbitmq.com/docs/heartbeats) (default: `30`)
  * `connection_timeout`: Bunny's socket open timeout (default: `11`)
  * `read_timeout`: Bunny's socket read timeout (default: `11`)
  * `write_timeout`: Bunny's socket write timeout (default: `11`)
@@ -534,7 +534,7 @@ Generate with
       <td>30</td>
       <td>Number</td>
       <td><tt>HUTCH_HEARTBEAT</tt></td>
-      <td><p><a href="http://rabbitmq.com/heartbeats.html">RabbitMQ heartbeat timeout</a></p></td>
+      <td><p><a href="https://www.rabbitmq.com/docs/heartbeats">RabbitMQ heartbeat timeout</a></p></td>
     </tr>
     <tr>
       <td><tt>channel_prefetch</tt></td>
@@ -590,7 +590,7 @@ Generate with
       <td>11</td>
       <td>Number</td>
       <td><tt>HUTCH_GRACEFUL_EXIT_TIMEOUT</tt></td>
-      <td><p>FIXME: DOCUMENT THIS</p></td>
+      <td><p>Timeout (in seconds) for consumer threads to finish before being killed during graceful shutdown</p></td>
     </tr>
     <tr>
       <td><tt>consumer_pool_size</tt></td>
@@ -646,7 +646,7 @@ Generate with
       <td>false</td>
       <td>Boolean</td>
       <td><tt>HUTCH_FORCE_PUBLISHER_CONFIRMS</tt></td>
-      <td><p>Enables publisher confirms, forces Hutch::Broker#wait_for_confirms for</p></td>
+      <td><p>Enables publisher confirms, forces Hutch::Broker#wait_for_confirms for every publish</p></td>
     </tr>
     <tr>
       <td><tt>enable_http_api_use</tt></td>
@@ -681,7 +681,7 @@ Generate with
       <td>''</td>
       <td>String</td>
       <td><tt>HUTCH_GROUP</tt></td>
-      <td></td>
+      <td><p>Consumer group to load (see Consumer Groups)</p></td>
     </tr>
   </tbody>
 </table>

data/bin/ci/before_build_docker.sh

@@ -1,11 +1,6 @@
 #!/bin/sh
 
 CTL=${MARCH_HARE_RABBITMQCTL:="docker exec rabbitmq rabbitmqctl"}
-PLUGINS=${MARCH_HARE_RABBITMQ_PLUGINS:="docker exec rabbitmq rabbitmq-plugins"}
-
-$PLUGINS enable rabbitmq_management
-
-sleep 3
 
 # guest:guest has full access to /
 
@@ -14,7 +9,7 @@ $CTL add_vhost /
 $CTL set_permissions -p / guest ".*" ".*" ".*"
 
 # Reduce retention policy for faster publishing of stats
-$CTL eval 'supervisor2:terminate_child(rabbit_mgmt_sup_sup, rabbit_mgmt_sup), application:set_env(rabbitmq_management,       sample_retention_policies, [{global, [{605, 1}]}, {basic, [{605, 1}]}, {detailed, [{10, 1}]}]), rabbit_mgmt_sup_sup:start_child().'
-$CTL eval 'supervisor2:terminate_child(rabbit_mgmt_agent_sup_sup, rabbit_mgmt_agent_sup), application:set_env(rabbitmq_management_agent, sample_retention_policies, [{global, [{605, 1}]}, {basic, [{605, 1}]}, {detailed, [{10, 1}]}]), rabbit_mgmt_agent_sup_sup:start_child().'
+$CTL eval 'supervisor2:terminate_child(rabbit_mgmt_sup_sup, rabbit_mgmt_sup), application:set_env(rabbitmq_management,       sample_retention_policies, [{global, [{605, 1}]}, {basic, [{605, 1}]}, {detailed, [{10, 1}]}]), rabbit_mgmt_sup_sup:start_child().' || true
+$CTL eval 'supervisor2:terminate_child(rabbit_mgmt_agent_sup_sup, rabbit_mgmt_agent_sup), application:set_env(rabbitmq_management_agent, sample_retention_policies, [{global, [{605, 1}]}, {basic, [{605, 1}]}, {detailed, [{10, 1}]}]), rabbit_mgmt_agent_sup_sup:start_child().' || true
 
 sleep 3

data/lib/hutch/acknowledgements/nack_on_all_failures.rb

@@ -10,7 +10,7 @@ module Hutch
         prefix = "message(#{properties.message_id || '-'}): "
         logger.debug "#{prefix} nacking message"
 
-        broker.nack(delivery_info.delivery_tag)
+        broker.nack(delivery_info.delivery_tag, channel: delivery_info.channel)
 
         true
       end

data/lib/hutch/adapters/bunny.rb

@@ -1,17 +1,21 @@
 require 'bunny'
 require 'forwardable'
 
+require 'hutch/logging'
+
 module Hutch
   module Adapters
     class BunnyAdapter
       extend Forwardable
+      include Logging
 
       DEFAULT_VHOST = Bunny::Session::DEFAULT_VHOST
 
       ConnectionRefused = Bunny::TCPConnectionFailed
+      ChannelAlreadyClosed = Bunny::ChannelAlreadyClosed
       PreconditionFailed = Bunny::PreconditionFailed
 
-      def_delegators :@connection, :start, :disconnect, :close, :create_channel, :open?
+      def_delegators :@connection, :start, :disconnect, :close, :create_channel, :open?, :recover_channel_topology
 
       def initialize(opts={})
         @connection = Bunny.new(opts)
@@ -25,6 +29,26 @@ module Hutch
         ch.prefetch(prefetch) if prefetch
       end
 
+      def queue_exists?(name)
+        @connection.queue_exists?(name)
+      end
+
+      def install_channel_recovery(ch)
+        # A consumer timeout invalidates the delivery tag, so a handler that
+        # was still running acknowledges an unknown tag and closes the channel.
+        ch.on_error do |channel, close|
+          next unless close.delivery_ack_timeout? || close.unknown_delivery_tag?
+
+          begin
+            channel.reopen
+            recover_channel_topology(channel)
+            logger.warn "recovered consumer channel closed with '#{close.reply_text}'"
+          rescue => ex
+            logger.error "channel recovery failed: #{ex.class}: #{ex.message}"
+          end
+        end
+      end
+
       def current_timestamp
         Time.now.to_i
       end

data/lib/hutch/adapters/march_hare.rb

@@ -9,6 +9,7 @@ module Hutch
       DEFAULT_VHOST = "/"
 
       ConnectionRefused = MarchHare::ConnectionRefused
+      ChannelAlreadyClosed = MarchHare::ChannelAlreadyClosed
       PreconditionFailed = MarchHare::PreconditionFailed
 
       def_delegators :@connection, :start, :disconnect, :close, :open?
@@ -25,6 +26,25 @@ module Hutch
         ch.prefetch = prefetch if prefetch
       end
 
+      # `MarchHare::Session` has no `queue_exists?` counterpart, and a failed
+      # passive declare closes the channel, hence the throwaway one.
+      def queue_exists?(name)
+        ch = @connection.create_channel
+        begin
+          ch.queue(name, passive: true)
+          true
+        rescue MarchHare::NotFound
+          false
+        ensure
+          ch.close rescue nil
+        end
+      end
+
+      # MarchHare::Channel has no on_error callback, and neither Channel#reopen
+      # nor Session#recover_channel_topology has a `march_hare` counterpart.
+      def install_channel_recovery(ch)
+      end
+
       def create_channel(n = nil, consumer_pool_size = 1, consumer_pool_abort_on_exception = false)
         @connection.create_channel(n)
       end
@@ -34,7 +54,7 @@ module Hutch
       end
 
       def self.new_exchange(ch, exchange_type, exchange_name, exchange_options)
-        MarchHare::Exchange.new(ch, exchange_name, exchange_options.merge(type: exchange_type))
+        MarchHare::Exchange.new(ch, exchange_name, exchange_options.merge(type: exchange_type)).tap(&:declare!)
       end
     end
   end