When a support request comes in at firstname.lastname@example.org we have a couple procedures to debug where a problem might come from.
Since it's good to share, here are a couple of our procedures.
Also see our list of known issues for a list of problems that might affect your application.
Our Ruby gem, Elixir package and Node.js package ship with a built-in diagnose command-line tool that outputs information about the configuration of the AppSignal library and environment it's running in. All this information can help in finding a potential cause of a problem.
If you open a support request, we'll usually ask you to run this first.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 # Ruby appsignal diagnose # Elixir mix appsignal.diagnose # With an Elixir release (`mix release`) binary: bin/your_app eval ":appsignal_tasks.diagnose()" # With a Distillery release binary: bin/your_app command appsignal_tasks diagnose # Node.js npx appsignal-diagnose
The diagnose command has to be run in the directory of an application that has the AppSignal library installed.
Unless the application is configured with environment variables it's necessary to provide the diagnose command with an environment. The environment will help AppSignal load the correct configuration that needs to be diagnosed.
1 2 3 4 5 6 7 8 9 10 11 12 # Ruby APPSIGNAL_APP_ENV=production appsignal diagnose # --environment option support since Ruby gem version 2.0.4 appsignal diagnose --environment=production # Elixir MIX_ENV=production mix appsignal.diagnose # Or with the release binary bin/your_app command appsignal_tasks diagnose # Node.js NODE_ENV=production npx appsignal-diagnose
The diagnose command will output the following data.
- Agent information
- Ruby gem / Elixir package / Node.js package version
- Agent version
- Ruby gem installation path
- C-extension / Nif loaded:
- Host information
- Hardware architecture
- Operating system
- Ruby / Elixir & OTP / Node.js version
- Heroku detection - only visible on Heroku
- Container detection
- Current user is
- Agent diagnostics
- Starts the agent in diagnose mode
- Agent configuration validation
- Agent logger initialization
- Agent lock file path writable check
- Push API key validation
- Tests if the Push API key that's being used is a valid key at AppSignal.com.
- Path information
- Tests all required paths if they are writable. Also outputs ownership of the current user.
current_path- path the diagnose command is run in. Should be a path for the application you're trying to debug. Usually the same as
root_path. (Ruby only)
root_path- application path. AppSignal will try to find a
config/appsignal.ymlfile here. (Ruby only)
log_dir_path- path the log file is created in.
log_file_path- path the log file is created. Is empty if no viable path could be found.
- Installation information (Ruby only)
- Something could have gone wrong during the installation of the AppSignal
agent. This section outputs the
mkmf.log(Makefile log) files.
- Something could have gone wrong during the installation of the AppSignal agent. This section outputs the
The following options need to be correctly configured for AppSignal to start. It's the absolute minimum, other configuration can also affect the instrumentation.
activeis set to
- The Push API key validates. An internet connection is required for this.
The configuration is key. It's important, because without it the AppSignal agent won't know which application it's instrumenting and in which environment.
Using the diagnose information we see if we can find a problem with the configuration of the agent.
The AppSignal agents have multiple methods of configuration.
We recommend you read the configuration topic to get started, specifically the minimal required configuration, configuration load order and the configuration options pages if you're experiencing problems.
- Configuration overview: Ruby / Elixir
- Minimal required configuration: Ruby / Elixir
- Configuration load order: Ruby / Elixir
- Configuration options: Ruby / Elixir
There can be many reasons why an application is not being detected by AppSignal. Only when the AppSignal servers start receiving data from an application it is created in the UI on AppSignal.com; if no data is received, no application is created.
When an integration is broken or not setup correctly it can take a long time to track the problem down. To rule out the AppSignal agent and its processing we can send "demo samples". (Note: this is currently not implemented for Node.js)
1 2 3 4 5 6 7 # Ruby appsignal demo --environment=development # Elixir MIX_ENV=prod mix appsignal.demo # Or with the release binary bin/your_app command appsignal_tasks demo
This command sends a fake web request and error sample to the AppSignal servers for the application. Once data has been received the AppSignal servers start processing the data and create an application on AppSignal.com.
In the AppSignal Ruby gem this process is also used in the
command-line tool to help with the installation process.
Note that the "demo" command has to be run in the directory of an application that has the AppSignal agent installed.
The "demo" command was added in Ruby gem version
When there's no problem found in the diagnose information, the agent's logs are the next thing you can look into. The AppSignal agents create log files to output useful information and problems that were encountered in the agent itself.
The AppSignal log file contains information from three different AppSignal components. It will prefix every log entry with a time stamp, component name, process PID, log level indicator and the log message itself.
The available agent components are:
Language specific implementation (Ruby / Elixir).
C-lang extension to the language implementation. Runs in the same process as
AppSignal Rust system agent. A separate process.
1 2 3 4 5 6 7 [2016-10-19T11:06:18 (process) #11329][DEBUG] Starting appsignal ^ ^ ^ ^ ^ | | | | Message | | | Log level | | Process PID | Agent component Timestamp in host time zone
For STDOUT loggers the AppSignal Ruby gem prefixes a recognizable "appsignal" prefix to the message so that specific AppSignal messages can be grepped in the parent process' log output.
1 2 3 4 5 6 7 8 [2016-10-19T11:06:18 (process) #11329][DEBUG] appsignal: Starting appsignal ^ ^ ^ ^ ^ ^ | | | | | Message | | | | AppSignal prefix | | | Log level | | Process PID | Agent component Timestamp in host time zone
There are two methods of saving logs from the AppSignal library. By writing the
logs to a log-file and by printing the logs in the process' STDOUT. See the
for more information.
Logs written to a log file are not written to your application's log files, but
have their own
appsignal.log file. Depending on the permissions of a host the
agent will write the logs to a different location.
The Ruby gem will first try to create a log file in an application's directory.
For Ruby on Rails applications it will create a log file in the
directory instead. If it has no luck creating a log file it will fall back on
/tmp directory. This is also where other AppSignal agent files
The Elixir package will write the log files to the
If it completely fails to find a writable location to save its logs to, it will
output them in the STDOUT of the parent application's process. This can also be
configured with the
The system agent is currently unable to log to STDOUT, so it will always log to
appsignal.log file even when the log is configured to STDOUT.
To let AppSignal tell you where it will write its logs to, see the output of the diagnose command.
On the Heroku hosting platform the integration library will log to STDOUT automatically. Since the agent can't log to STDOUT it will continue to write to the
On Heroku it's not possible to see the log file contents with
heroku run bash and then read the log file, because of the way Heroku Dynos are containerized.
Heroku provides an add-on called "Heroku Exec" to allow executing commands in the same dyno container as your application. Using this add-on we can read the log file on your application's dynos.
1 2 3 4 # Install the Exec add-on as described on the Heroku website: # https://devcenter.heroku.com/articles/heroku-exec $ heroku ps:exec $ cat /path/to/appsignal.log
When an application with AppSignal integration starts it boots the AppSignal system agent. This agent runs as a separate process and communicates with the application. After processing instrumentation data, it is this agent that sends the data to the AppSignal servers, not the language specific agent. When the system agent is not running, no data can be processed or sent to AppSignal.
It's important to know if the system agent is running. You can find it in the
process list under the name
1 ps aux | grep appsignal
ps-command on the command-line on your host to see if the agent is
running while your application is running AppSignal.
There can be many factors at play when a problem occurs. Things to check when AppSignal doesn't seem to be working or there are no logs available.
- Since when does the problem occur?
- Was the agent updated?
- Were other libraries updated?
- What does the AppSignal agent logs say with log-level "debug"?
- Is the Operating System supported?
- Did the "extension" install and load correctly?
Answers should be in the "diagnose" output.
- Is the application running inside a containerized system?
- Are the application servers's synced using NTP to prevent clock drift?
- Is your problem mentioned in our list of known issues?
If the steps described above haven't shown a cause for the problem, a good next step is to try to replicate the situation in the most minimal setup possible.
Create a test application with as little configuration as possible, loading only the minimum required libraries and dependencies.
With this reproducible example application we can start tracking down the cause of the problem.
Don't hesitate to contact us if you run into any issues while debugging the AppSignal agents. We're here to help.