diff --git a/docs/creating-alerters.rst b/docs/creating-alerters.rst new file mode 100644 index 00000000..729f1f46 --- /dev/null +++ b/docs/creating-alerters.rst @@ -0,0 +1,68 @@ +Creating Alerters +================= + +To create your own Alerter, you need to: + +1. Create a Python file in :file:`simplemonitor/Alerters` (or pick a suitable existing one to add it to) +2. If you're creating a new file, you'll need a couple of imports: + + .. code-block:: python + + from ..Monitors.monitor import Monitor + from .alerter import Alerter, AlertLength, AlertType, register + +3. Define your alerter class, which should subclass ``Alerter`` and be decorated by ``@register``. Set a class attribute for the "type" which will be used in the alerter configuration to use it. + + .. code-block:: python + + @register + class MyAlerter(Alerter): + + alerter_type = "my_alerter" + +4. Define your initialiser. It should call the superclass's initialiser, and then read its configuration values from the supplied dict. You can also do any other initialisation here. + + This code should be safe to re-run, as if SimpleMonitor reloads its configuration, it will call ``__init__()`` with the new configuration dict. Use the :py:func:`get_config_option` helper to read config values. + + .. code-block:: python + + @register + class MyAlerter(Alerter): + + alerter_type = "my_alerter" + + def __init__(self, config_options: dict) -> None: + super().__init__(config_options) + self.my_setting = self.get_config_option("setting", required=True) + +5. Add a ``send_alerter`` function. This receives the information for a single monitor. You should first call ``self.should_alert(monitor)``, which will return the type of alert to be sent (e.g. failure). You should return if it is ``AlertType.NONE``. + + You should then prepare your message. Call ``self.build_message()`` to generate the message content. Check the value of ``self._dry_run`` and if it is True, you should log (using ``self.alerter_logger.info(...)``) what you would do, else you should do it. + + .. py:function:: Alerter.build_message(length: AlertLength, alert_type: AlertType, monitor: Monitor) -> str + + Generate a suitable length alert message for the given type of alert, for the given Monitor. + + :param AlertLength: one of the AlertLength enum values: ``NOTIFICATION`` (shortest), ``SMS`` (will be <= 140 chars), ``ONELINE``, ``TERSE`` (not currently supported), ``FULL``, or ``ESSAY`` + :param AlertType: one of the AlertType enum values: ``NONE``, ``FAILURE``, ``CATCHUP``, or ``SUCCESS`` + :param monitor: the Monitor to generate the message for + :return: the built message + :rtype: str + :raises ValueError: if the AlertType is unknown + :raises NotImplementedError: if the AlertLength is unknown or unsupported + +7. You should also give a ``_describe_action`` function, which explains what this alerter does. Note that the time configuration for the alerter will be automatically added: + + .. code-block:: python + + @register + class MyAlerter(Alerter): + + # ... + + def _describe_action(self) -> str: + return f"sending FooAlerters to {self.recipient}" + +7. In :file:`simplemonitor/Alerters/__init__.py`, add your Alerter to the list of imports. + +That's it! You should now be able to use ``type=my_alerter`` in your Alerters configuration to use your alerter. diff --git a/docs/creating-loggers.rst b/docs/creating-loggers.rst new file mode 100644 index 00000000..564729bf --- /dev/null +++ b/docs/creating-loggers.rst @@ -0,0 +1,95 @@ +Creating Loggers +================ + +Before writing your logger, you need to consider if you should support **batching** or not. If a logger supports batching, then it collects all the monitor results and then performs its logging action. For example, the HTML logger uses batching so that when it generates the HTML output, it knows all the monitors to include (and can sort them etc). Non-batching loggers will simply perform their logging action multiple times, once per monitor. + +To create your own Logger, you need to: + +1. Create a Python file in :file:`simplemonitor/Loggers` (or pick a suitable existing one to add it to) +2. If you're creating a new file, you'll need a couple of imports: + + .. code-block:: python + + from ..Monitors.monitor import Monitor + from .logger import Logger, register + +3. Define your logger class, which should subclass ``Logger`` and be decorated by ``@register``. Set a class attribute for the "type" which will be used in the logger configuration to use it. Additionally, set the ``supports_batch`` value to indicate if your logger should be used in batching mode. + + .. code-block:: python + + @register + class MyLogger(Logger): + + logger_type = "my_logger" + supports_batch = True # or False + +4. Define your initialiser. It should call the superclass's initialiser, and then read its configuration values from the supplied dict. You can also do any other initialisation here. + + This code should be safe to re-run, as if SimpleMonitor reloads its configuration, it will call ``__init__()`` with the new configuration dict. Use the :py:func:`get_config_option` helper to read config values. + + .. code-block:: python + + @register + class MyLogger(Logger): + + logger_type = "my_logger" + + def __init__(self, config_options: dict) -> None: + super().__init__(config_options) + self.my_setting = self.get_config_option("setting", required=True) + +5. Add a ``save_result2`` function (yes, I know). This receives the information for a single monitor. + + **Batching loggers** should save the information they need to into `self.batch_data`, which should (but does not have to be) a dict of `str: Any` using the monitor name as the key. This is automatically initialised to an empty dict at the start of the batch. You should extend the `start_batch` method from `Logger` to customise it. + + .. code-block:: python + + @register + class MyLogger(Logger): + + # ... + + def save_result2(self, name: str, monitor: Monitor) -> None: + self.batch_data[name] = monitor.state + + **Non-batching loggers** can perform whatever logging action they are designed for at this point. + + .. code-block:: python + + @register + class MyLogger(Logger): + + # ... + + def save_result2(self, name: str, monitor: Monitor) -> None: + self._my_logger_action(f"Monitor {name} is in state {monitor.state}") + +6. **Batching loggers** only should provide a ``process_batch`` method, which is called after all the monitors have been processed. This is where you should perform your batched logging operation. + + .. code-block:: python + + @register + class MyLogger(Logger): + + # ... + + def process_batch(self) -> None: + with open(self.filename, "w") as file_handle: + for monitor, state in self.batch_data.iteritems(): + file_handle.write(f"Monitor {monitor} is in state {state}\n") + +7. You should also give a ``describe`` function, which explains what this logger does: + + .. code-block:: python + + @register + class MyLogger(Logger): + + # ... + + def describe(self) -> str: + return f"writing monitor info to {self.filename}" + +7. In :file:`simplemonitor/Loggers/__init__.py`, add your Logger to the list of imports. + +That's it! You should now be able to use ``type=my_thing`` in your Loggers configuration to use your logger.