AMQP
Simple Elixir wrapper for the Erlang RabbitMQ client.
The API is based on Langohr, a Clojure client for RabbitMQ.
Migration from 0.X to 1.X
If you use amqp 0.X and plan to migrate to 1.0 please read our migration guide.
Usage
Add AMQP as a dependency in your mix.exs
file.
def deps do
[{:amqp, "~> 1.2"}]
end
Elixir will start amqp
automatically with this if you use Elixir 1.6+.
If that's not the case (use Application.started_applications/0
to check), try adding :amqp
to applications
or extra_applications
in your mix.exs
. Or call Application.ensure_started(:amqp)
at the start.
After you are done, run mix deps.get
in your shell to fetch and compile AMQP. Start an interactive Elixir shell with iex -S mix
.
iex> {:ok, conn} = AMQP.Connection.open()
# {:ok, %AMQP.Connection{pid: #PID<0.165.0>}}
iex> {:ok, chan} = AMQP.Channel.open(conn)
# {:ok, %AMQP.Channel{conn: %AMQP.Connection{pid: #PID<0.165.0>}, pid: #PID<0.177.0>}
iex> AMQP.Queue.declare(chan, "test_queue")
# {:ok, %{consumer_count: 0, message_count: 0, queue: "test_queue"}}
iex> AMQP.Exchange.declare(chan, "test_exchange")
# :ok
iex> AMQP.Queue.bind(chan, "test_queue", "test_exchange")
# :ok
iex> AMQP.Basic.publish(chan, "test_exchange", "", "Hello, World!")
# :ok
iex> {:ok, payload, meta} = AMQP.Basic.get(chan, "test_queue")
iex> payload
# "Hello, World!"
iex> AMQP.Queue.subscribe(chan, "test_queue", fn payload, _meta -> IO.puts("Received: #{payload}") end)
# {:ok, "amq.ctag-5L8U-n0HU5doEsNTQpaXWg"}
iex> AMQP.Basic.publish(chan, "test_exchange", "", "Hello, World!")
# :ok
# Received: Hello, World!
Setup a consumer GenServer
defmodule Consumer do
use GenServer
use AMQP
def start_link do
GenServer.start_link(__MODULE__, [], [])
end
@exchange "gen_server_test_exchange"
@queue "gen_server_test_queue"
@queue_error "#{@queue}_error"
def init(_opts) do
{:ok, conn} = Connection.open("amqp://guest:guest@localhost")
{:ok, chan} = Channel.open(conn)
setup_queue(chan)
# Limit unacknowledged messages to 10
:ok = Basic.qos(chan, prefetch_count: 10)
# Register the GenServer process as a consumer
{:ok, _consumer_tag} = Basic.consume(chan, @queue)
{:ok, chan}
end
# Confirmation sent by the broker after registering this process as a consumer
def handle_info({:basic_consume_ok, %{consumer_tag: consumer_tag}}, chan) do
{:noreply, chan}
end
# Sent by the broker when the consumer is unexpectedly cancelled (such as after a queue deletion)
def handle_info({:basic_cancel, %{consumer_tag: consumer_tag}}, chan) do
{:stop, :normal, chan}
end
# Confirmation sent by the broker to the consumer process after a Basic.cancel
def handle_info({:basic_cancel_ok, %{consumer_tag: consumer_tag}}, chan) do
{:noreply, chan}
end
def handle_info({:basic_deliver, payload, %{delivery_tag: tag, redelivered: redelivered}}, chan) do
# You might want to run payload consumption in separate Tasks in production
consume(chan, tag, redelivered, payload)
{:noreply, chan}
end
defp setup_queue(chan) do
{:ok, _} = Queue.declare(chan, @queue_error, durable: true)
# Messages that cannot be delivered to any consumer in the main queue will be routed to the error queue
{:ok, _} = Queue.declare(chan, @queue,
durable: true,
arguments: [
{"x-dead-letter-exchange", :longstr, ""},
{"x-dead-letter-routing-key", :longstr, @queue_error}
]
)
:ok = Exchange.fanout(chan, @exchange, durable: true)
:ok = Queue.bind(chan, @queue, @exchange)
end
defp consume(channel, tag, redelivered, payload) do
number = String.to_integer(payload)
if number <= 10 do
:ok = Basic.ack channel, tag
IO.puts "Consumed a #{number}."
else
:ok = Basic.reject channel, tag, requeue: false
IO.puts "#{number} is too big and was rejected."
end
rescue
# Requeue unless it's a redelivered message.
# This means we will retry consuming a message once in case of exception
# before we give up and have it moved to the error queue
#
# You might also want to catch :exit signal in production code.
# Make sure you call ack, nack or reject otherwise comsumer will stop
# receiving messages.
exception ->
:ok = Basic.reject channel, tag, requeue: not redelivered
IO.puts "Error converting #{payload} to integer"
end
end
iex> Consumer.start_link
{:ok, #PID<0.261.0>}
iex> {:ok, conn} = AMQP.Connection.open
{:ok, %AMQP.Connection{pid: #PID<0.165.0>}}
iex> {:ok, chan} = AMQP.Channel.open(conn)
{:ok, %AMQP.Channel{conn: %AMQP.Connection{pid: #PID<0.165.0>}, pid: #PID<0.177.0>}
iex> AMQP.Basic.publish chan, "gen_server_test_exchange", "", "5"
:ok
Consumed a 5.
iex> AMQP.Basic.publish chan, "gen_server_test_exchange", "", "42"
:ok
42 is too big and was rejected.
iex> AMQP.Basic.publish chan, "gen_server_test_exchange", "", "Hello, World!"
:ok
Error converting Hello, World! to integer
Error converting Hello, World! to integer
Stable RabbitMQ Connection
While the above example works, it does nothing to handle RabbitMQ connection outages. In case of an outage your Genserver will remain stale and won't receive any messages from the broker as the connection is never restarted.
Luckily, implementing a reconnection logic is quite straight forward. Since the connection record holds the pid of the connection itself, we can monitor it and get a notification when it goes down.
Example implementation:
defmodule MyApp.AMQP do
use GenServer
require Logger
alias AMQP.Connection
@host "amqp://localhost"
@reconnect_interval 10_000
def start_link(opts \\ [name: __MODULE__]) do
GenServer.start_link(__MODULE__, nil, opts)
end
def init(_) do
send(self(), :connect)
{:ok, nil}
end
def get_connection do
case GenServer.call(__MODULE__, :get) do
nil -> {:error, :not_connected}
conn -> {:ok, conn}
end
end
def handle_call(:get, _, conn) do
{:reply, conn, conn}
end
def handle_info(:connect, conn) do
case Connection.open(@host) do
{:ok, conn} ->
# Get notifications when the connection goes down
Process.monitor(conn.pid)
{:noreply, conn}
{:error, _} ->
Logger.error("Failed to connect #{@host}. Reconnecting later...")
# Retry later
Process.send_after(self(), :connect, @reconnect_interval)
{:noreply, nil}
end
end
def handle_info({:DOWN, _, :process, _pid, reason}, _) do
# Stop GenServer. Will be restarted by Supervisor.
{:stop, {:connection_lost, reason}, nil}
end
end
Now, when the server starts, it will try to reconnect indefinitely until it succeeds. When the connection drops or the server is down, the GenServer will stop. If you have put the GenServer module to your application tree, the Supervisor will automatically restart it. Then it will try to reconnect indefinitely until it succeeds.
Types of arguments and headers
The parameter arguments
in Queue.declare
, Exchange.declare
, Basic.consume
and the parameter headers
in Basic.publish
are a list of tuples in the form {name, type, value}
, where name
is a binary containing the argument/header name, type
is an atom describing the AMQP field type and value
a term compatible with the AMQP field type.
The valid AMQP field types are:
:longstr
| :signedint
| :decimal
| :timestamp
| :table
| :byte
| :double
| :float
| :long
| :short
| :bool
| :binary
| :void
| :array
Valid argument names in Queue.declare
include:
- "x-expires"
- "x-message-ttl"
- "x-dead-letter-routing-key"
- "x-dead-letter-exchange"
- "x-max-length"
- "x-max-length-bytes"
Valid argument names in Basic.consume
include:
- "x-priority"
- "x-cancel-on-ha-failover"
Valid argument names in Exchange.declare
include:
- "alternate-exchange"
Troubleshooting / FAQ
Connections and Channels
If this is your first time using RabbitMQ, we recommend you to start designing your application like this way:
- Open and manage a single connection for an application
- Open/close a channel per process (don't share a channel between multiple processes)
Once you saw things in action you can now consider optimising the performance by increasing number of connections etc.
Note it's completely safe to share a single connection between multiple processes. However it is not recommended to share a channel between multiple processes. It's technically possible but you want to understand the implications when you do.
Make sure you close the channel after used to avoid any potential memory leaks and warnings from RabbitMQ client library.
Consumer stops receiving messages
It usually happens when your code doesn't send acknowledgement(ack, nack or reject) after receiving a message. You want to investigate if...
- an exception was raised and how it would be handled
- :exit signal was thrown and how it would be handled
- a message processing took long time.
If you use GenServer in consumer, try storing number of messages the server is
currently processing to the GenServer state.
If the number equals prefetch_count
, those messages were left without
acknowledgements and that's why consumer have stopped receiving more
messages.
Old version of Elixir or OTP
OTP 17 and 18 are supported only on version 0.1.x. Please understand that we won't make further changes to 0.1 except for major security issues.
Heartbeats
In case the connection is dropped automatically, consider enabling heartbeats.
You can set heartbeat
option when you open a connection.
For more details, read this article
Log related to amqp supervisors are too verbose
Try the following configuration.
config :logger, handle_otp_reports: false
Or try filtering out the messages at your application start:
:logger.add_primary_filter(
:ignore_rabbitmq_progress_reports,
{&:logger_filters.domain/2, {:stop, :equal, [:progress]}}
)
See this comment for the details.
Lager conflicts with Elixir logger
Lager is used by rabbit_common and it is not Elixir's best friend yet. You need a workaround.
In mix.exs, you have to load :lager before :logger.
extra_applications: [:lager, :logger, :amqp]
Here is a sample configuration to silent rabbit_common logging.
config :lager,
error_logger_redirect: false,
handlers: [level: :critical]
Check out Lager and RabbitMQ documentation for more information.
Compile error on ranch_proxy_protocol with OTP 21
Update amqp to 1.1.0 or a greater version.
Does the library support AMQP 1.0?
Currently the library doesn't support AMQP 1.0 and there is no plan to do so at the moment. Our main aim here (at least for now) is to provide a thin wrapper around amqp_client for Elixir programmers.