The Official Sentry Client for Elixir which provides a simple API to capture exceptions, automatically handle Plug Exceptions and provides a backend for the Elixir Logger.
To use Sentry with your projects, edit your mix.exs file to add it as a dependency and add the :sentry
package to your applications:
defp application do
[applications: [:sentry, :logger]]
end
defp deps do
[{:sentry, "~> 6.4"}]
end
Sometimes you want to capture specific exceptions. To do so, use Sentry.capture_exception/2
.
try do
ThisWillError.reall()
rescue
my_exception ->
Sentry.capture_exception(my_exception, [stacktrace: System.stacktrace(), extra: %{extra: information}])
end
Sometimes you want to capture arbitrary messages that are not Exceptions.
Sentry.capture_message("custom_event_name", extra: %{extra: information})
For optional settings check the docs.
In your Plug.Router or Phoenix.Router, add the following lines:
use Plug.ErrorHandler
use Sentry.Plug
If you are using Phoenix, you can also include Sentry.Phoenix.Endpoint in your Endpoint. This module captures errors occurring in the Phoenix pipeline before the request reaches the Router:
use Phoenix.Endpoint, otp_app: :my_app
use Sentry.Phoenix.Endpoint
More information on why this may be necessary can be found here: getsentry#229 and phoenixframework/phoenix#2791
This library comes with an extension to capture all error messages that the Plug handler might not. This is based on the Erlang error_logger.
To set this up, add :ok = :error_logger.add_report_handler(Sentry.Logger)
to your application's start function. Example:
def start(_type, _opts) do
children = [
supervisor(MyApp.Repo, []),
supervisor(MyAppWeb.Endpoint, [])
]
opts = [strategy: :one_for_one, name: MyApp.Supervisor]
:ok = :error_logger.add_report_handler(Sentry.Logger)
Supervisor.start_link(children, opts)
end
Key | Required | Default | Notes |
---|---|---|---|
dsn |
True | n/a | |
environment_name |
False | :dev |
|
included_environments |
False | [:test, :dev, :prod] |
If you need non-standard mix env names you need to include it here |
tags |
False | %{} |
|
release |
False | None | |
server_name |
False | None | |
client |
False | Sentry.Client |
If you need different functionality for the HTTP client, you can define your own module that implements the Sentry.HTTPClient behaviour and set client to that module |
hackney_opts |
False | [pool: :sentry_pool] |
|
hackney_pool_max_connections |
False | 50 | |
hackney_pool_timeout |
False | 5000 | |
before_send_event |
False | ||
after_send_event |
False | ||
sample_rate |
False | 1.0 | |
in_app_module_whitelist |
False | [] |
|
report_deps |
False | False | Will attempt to load Mix dependencies at compile time to report alongside events. This was switched to default to false following a bug in the parallel compiler (#232 / elixir-lang/elixir#7699) |
enable_source_code_context |
False | False | |
root_source_code_path |
Required if enable_source_code_context is enabled |
Should generally be set to File.cwd! |
|
context_lines |
False | 3 | |
source_code_exclude_patterns |
False | [~r"/_build/", ~r"/deps/", ~r"/priv/"] |
|
source_code_path_pattern |
False | "**/*.ex" |
|
filter |
False | Module where the filter rules are defined (see Filtering Exceptions) |
An example production config might look like this:
config :sentry,
dsn: "https://public_key@app.getsentry.com/1",
environment_name: :prod,
included_environments: [:prod],
enable_source_code_context: true,
root_source_code_path: File.cwd!,
tags: %{
env: "production"
},
hackney_opts: [pool: :my_pool],
in_app_module_whitelist: [MyApp]
The environment_name
and included_environments
work together to determine
if and when Sentry should record exceptions. The environment_name
is the
name of the current environment. In the example above, we have explicitly set
the environment to :prod
which works well if you are inside an environment
specific configuration like config/prod.exs
.
Alternatively, you could use Mix.env in your general configuration file:
config :sentry, dsn: "https://public_key@app.getsentry.com/1",
included_environments: [:prod],
environment_name: Mix.env
You can even rely on more custom determinations of the environment name. It's not uncommon for most applications to have a "staging" environment. In order to handle this without adding an additional Mix environment, you can set an environment variable that determines the release level.
config :sentry, dsn: "https://public_key@app.getsentry.com/1",
included_environments: ~w(production staging),
environment_name: System.get_env("RELEASE_LEVEL") || "development"
In this example, we are getting the environment name from the RELEASE_LEVEL
environment variable. If that variable does not exist, we default to "development"
.
Now, on our servers, we can set the environment variable appropriately. On
our local development machines, exceptions will never be sent, because the
default value is not in the list of included_environments
.
Sentry uses the hackney HTTP client for HTTP requests. Sentry starts its own hackney pool named :sentry_pool
with a default connection pool of 50, and a connection timeout of 5000 milliseconds. The pool can be configured with the hackney_pool_max_connections
and hackney_pool_timeout
configuration keys. If you need to set other hackney configurations for things like a proxy, using your own pool or response timeouts, the hackney_opts
configuration is passed directly to hackney for each request.
Sentry has multiple options for including contextual information. They are organized into "Tags", "User", and "Extra", and Sentry's documentation on them is here. Breadcrumbs are a similar concept and Sentry's documentation covers them here.
In Elixir this can be complicated due to processes being isolated from one another. Tags context can be set globally through configuration, and all contexts can be set within a process, and on individual events. If an event is sent within a process that has some context configured it will include that context in the event. Examples of each are below, and for more information see the documentation of Sentry.Context.
# Global Tags context via configuration:
config :sentry,
tags: %{my_app_version: "14.30.10"}
# ...
# Process-based Context
Sentry.Context.set_extra_context(%{day_of_week: "Friday"})
Sentry.Context.set_user_context(%{id: 24, username: "user_username", has_subscription: true})
Sentry.Context.set_tags_context(%{locale: "en-us"})
Sentry.Context.add_breadcrumb(%{category: "web.request"})
# Event-based Context
Sentry.capture_exception(exception, [tags: %{locale: "en-us", }, user: %{id: 34},
extra: %{day_of_week: "Friday"}, breadcrumbs: [%{timestamp: 1461185753845, category: "web.request"}]]
By default, Sentry aggregates reported events according to the attributes of the event, but users may need to override this functionality via fingerprinting.
To achieve that in Sentry Elixir, one can use the before_send_event
configuration callback. If there are certain types of errors you would like to have grouped differently, they can be matched on in the callback, and have the fingerprint attribute changed before the event is sent. An example configuration and implementation could look like:
# lib/sentry.ex
defmodule MyApp.Sentry
def before_send(%{exception: [%{type: DBConnection.ConnectionError}]} = event) do
%{event | fingerprint: ["ecto", "db_connection", "timeout"]}
end
def before_send(event) do
event
end
end
# config.exs
config :sentry,
before_send_event: {MyApp.Sentry, :before_send},
# ...
Sentry's server supports showing the source code that caused an error, but depending on deployment, the source code for an application is not guaranteed to be available while it is running. To work around this, the Sentry library reads and stores the source code at compile time. This has some unfortunate implications. If a file is changed, and Sentry is not recompiled, it will still report old source code.
The best way to ensure source code is up to date is to recompile Sentry itself via mix deps.compile sentry --force
. It's possible to create a Mix Task alias in mix.exs
to do this. The example below would allow one to run mix sentry_recompile
which will force recompilation of Sentry so it has the newest source and then compile the project:
# mix.exs
defp aliases do
[sentry_recompile: ["deps.compile sentry --force", "compile"]]
end
For more documentation, see Sentry.Sources.
To ensure you've set up your configuration correctly we recommend running the included mix task. It can be tested on different Mix environments and will tell you if it is not currently configured to send events in that environment:
$ MIX_ENV=dev mix sentry.send_test_event
Client configuration:
server: https://sentry.io/
public_key: public
secret_key: secret
included_environments: [:prod]
current environment_name: :dev
:dev is not in [:prod] so no test event will be sent
$ MIX_ENV=prod mix sentry.send_test_event
Client configuration:
server: https://sentry.io/
public_key: public
secret_key: secret
included_environments: [:prod]
current environment_name: :prod
Sending test event!
To build the docs locally, you'll need the Sphinx:
$ pip install sphinx
Once Sphinx is available building the docs is simply:
$ make docs
You can then view the docs in your browser:
$ open docs/_build/html/index.html
In some cases, users may want to test that certain actions in their application cause a report to be sent to Sentry. Sentry itself does this by using Bypass. It is important to note that when modifying the environment configuration the test case should not be run asynchronously. Not returning the environment configuration to its original state could also affect other tests depending on how the Sentry configuration interacts with them.
Example:
test "add/2 does not raise but sends an event to Sentry when given bad input" do
bypass = Bypass.open()
Bypass.expect(bypass, fn conn ->
{:ok, _body, conn} = Plug.Conn.read_body(conn)
Plug.Conn.resp(conn, 200, ~s<{"id": "340"}>)
end)
Application.put_env(:sentry, :dsn, "http://public:secret@localhost:#{bypass.port}/1")
MyModule.add(1, "a")
end
This project is Licensed under the MIT License.