prometheus-alert-model 0.1.0

Pydantic model of the Prometheus Alertmanager alert payload

Prometheus Alert Model for Python

This project provides the Pydantic models AlertGroup and Alert that represent a payload from Prometheus Alertmanager. In addition, it also includes a number of useful utility methods that perform actions on alert data. It can be used as a drop-in wherever you want to work with Prometheus Alertmanager alert payload data.

A prominent example for using it is a FastAPI route that receives alert payloads. Simply add AlertGroup as a parameter to the handler.

Features

• Pydantic models that matches the official Alertmanager payload schema.
• Fields specific_annotations and specific_labels in every alert that contain elements that are specific to the respective alert.
• Methods to update common annotations and labels.
• Methods to remove, add, update, override and prefix annotations and labels.
• Every single method is well covered by tests.

Table of Contents

• Usage
• Motivation
• Development

Usage

Using the model is pretty straight-forward. Take a look at the automatically generated docs or the source code itself.

Here is a very short example how you could use the model. It removes all annotations and labels starting with two underscores and adds a prefix that contains namespace info to the summary label. Finally it prints specific elements for all alerts in the group.

from prometheus_alert_model import AlertGroup
from fastapi import FastAPI
from re import compile

app = FastAPI()

@app.post("/alert")
def post_alert(alert_group: AlertGroup):
    alert_group.remove_re(
        annotations=r"^(__.*)$",
        labels=compile(r"^(__.*)$")
    )

    alert_group.add_prefix(labels={"summary": "Prototyping system: "})
    
    for alert in alert_group.alerts:
        print(alert.specific_annotations
        print(alert.specific_labels)

Alert Model

In the following all attributes you can find within a Alert. Notice the custom attributes specific_annotations and specific_labels that include elements that are specific to the respective alert in context of the complete AlertGroup / payload.

fingerprint: str
status: str
starts_at: datetime
ends_at: datetime
generator_url: str
annotations: Dict[str, str]
labels: Dict[str, str]
specific_annotations: Dict[str, str]
specific_labels: Dict[str, str]

AlertGroup Model

In the following all attributes you can find within a AlertGroup. It represents a single payload from Alertmanager.

receiver: str
status: str
external_url: str = Field(alias="externalURL")
version: str
group_key: str = Field(alias="groupKey")
truncated_alerts: int = Field(alias="truncatedAlerts", default=0)
group_labels: Dict[str, str] = Field(alias="groupLabels")
common_annotations: Dict[str, str] = Field(alias="commonAnnotations")
common_labels: Dict[str, str] = Field(alias="commonLabels")
alerts: List[Alert]

Here is a short summary over the included utility methods (for full documentation please refer to type hints or the automatically generated docs):

• update_specific_elements: Updates specific labels and annotations.
• update_specific_annotations: Updates specific annotations.
• update_specific_labels: Updates specific labels.
• update_common_elements: Updates common annotations and labels.
• update_common_annotations: Updates common annotations.
• update_common_labels: Updates common labels.
• remove: Removes annotations and labels by name.
• remove_re: Removes annotations and labels by matching names with regex.
• add: Adds annotations and labels but skips existing elements.
• override: Adds annotations and labels and overrides existing elements.
• add_prefix: Adds prefix to annotations and labels.

Motivation

I have a bunch of Python scripts that work in some shape or form with Prometheus Alertmanager data. Instead of duplicating the model across all of them I prefer to have a single small package that is well tested and reuse it again and again. This way I don't have to reimplement utility functions / methods.

Development

Please refer to "DEVELOPMENT.md".