adding PagerDuty support

.gitignore

@@ -7,3 +7,4 @@ venv
 *.egg-info
 *.snap
 **/__pycache__
+*.rock

CHANGELOG.md

@@ -7,6 +7,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.9.0] - 2024-05-30
+
+- Adding PagerDuty native support (#76).
+
+
 ## [0.8.0] - 2024-03-07
 
 - Fixes container silently running by exiting with non-zero status when configuration file is missing. (#70).

CONTRIBUTING.md

@@ -28,6 +28,23 @@ cp config-defaults.yaml cos-alerter.yaml
 docker run -p 8080:8080 --rm --mount type=bind,source="$(pwd)"/cos-alerter.yaml,target=/etc/cos-alerter.yaml,readonly -it cos-alerter:0.2.0
 ```
 
+## Run With Kubernetes
+Prepare the image:
+```shell
+rockcraft pack
+# update <registry-ip> with the actual IP of your docker registry
+# update <image-tag> with the image tag you would like to use in testing
+skopeo copy oci-archive:cos-alerter_0.8.0_amd64.rock docker://<registry-ip>/<image-tag> --dest-tls-verify=false
+```
+
+Run:
+```shell
+# in k8s-local-test/deploy.yaml update <registry-ip>/<image-tag> with appropriate values
+# in k8s-local-test/deploy.yaml update cos-alerter.yaml configmap with appropriate values
+kubectl apply -f k8s-local-test/deploy.yaml
+kubectl apply -f k8s-local-test/svc.yaml
+```
+
 ## Run Tests
 
 * `pip install tox`

cos_alerter/alerter.py

@@ -13,10 +13,12 @@
 import time
 import typing
 from pathlib import Path
+from typing import Optional
 
 import apprise
 import durationpy
 import xdg_base_dirs
+from pdpyras import EventsAPISession
 from ruamel.yaml import YAML
 from ruamel.yaml.constructor import DuplicateKeyError
 
@@ -211,6 +213,14 @@ def clients():
 
     def reset_alert_timeout(self):
         """Set the "last alert time" to right now."""
+        # In case an instance was down, resolve the PagerDuty incident before resetting the last alert time
+        if self.is_down():
+            pagerduty_destinations, _ = split_destinations()
+            handle_pagerduty_incidents(
+                incident_type="resolve",
+                dedup_key=f"{self.clientid}-{self.last_alert_datetime()}",
+                destinations_list=pagerduty_destinations,
+            )
         logger.debug("Resetting alert timeout for %s.", self.clientid)
         self.data["alert_time"] = time.monotonic()
 
@@ -274,7 +284,12 @@ def notify(self):
         # Sending notifications can be a long operation so handle that in a separate thread.
         # This avoids interfering with the execution of the main loop.
         notify_thread = threading.Thread(
-            target=send_notifications, kwargs={"title": title, "body": body}
+            target=send_notifications,
+            kwargs={
+                "title": title,
+                "body": body,
+                "dedup_key": f"{self.clientid}-{self.last_alert_datetime()}",
+            },
         )
         notify_thread.start()
 
@@ -290,20 +305,62 @@ def up_time():
     return time.monotonic() - state["start_time"]
 
 
-def send_notifications(title: str, body: str):
+def split_destinations():
+    """Split destinations into PagerDuty and non-PagerDuty lists."""
+    pagerduty_destinations = [
+        source for source in config["notify"]["destinations"] if source.startswith("pagerduty")
+    ]
+    non_pagerduty_destinations = [
+        source for source in config["notify"]["destinations"] if not source.startswith("pagerduty")
+    ]
+    return pagerduty_destinations, non_pagerduty_destinations
+
+
+def send_notifications(title: str, body: str, dedup_key: str):
     """Send a notification to all receivers."""
     # TODO: Since this is run in its own thread, we have to make sure we properly
     # log failures here.
+    pagerduty_destinations, non_pagerduty_destinations = split_destinations()
+
+    # Send notifications to non-PagerDuty destinations
     sender = apprise.Apprise()
-    for source in config["notify"]["destinations"]:
+    for source in non_pagerduty_destinations:
         sender.add(source)
     sender.notify(title=title, body=body)
 
+    # Send notifications to PagerDuty destinations
+    handle_pagerduty_incidents("trigger", dedup_key, pagerduty_destinations, body)
+
+
+def handle_pagerduty_incidents(
+    incident_type: str,
+    dedup_key: str,
+    destinations_list: list,
+    incident_summary: Optional[str] = None,
+):
+    """Handles PagerDuty incidents by triggering or resolving incidents based on the specified incident type.
+
+    Args:
+        incident_type (str): The type of incident action to perform. Should be either 'trigger' or 'resolve'.
+        dedup_key (str): The deduplication key to uniquely identify the incident.
+        destinations_list (list): List of destinations to handle PagerDuty incidents for.
+        incident_summary (str, optional): A summary of the incident, used only when triggering an incident. Defaults to None.
+    """
+    for source in destinations_list:
+        integration_key = source.split("//")[1].split("@")[0]
+        session = EventsAPISession(integration_key)
+
+        if incident_type == "trigger":
+            session.trigger(source="cos-alerter", summary=incident_summary, dedup_key=dedup_key)
+        elif incident_type == "resolve":
+            session.resolve(dedup_key)
+
 
 def send_test_notification():
     """Signal handler which sends a test email to all configured receivers."""
     logger.info("Sending test notifications.")
     send_notifications(
         title="COS-Alerter test email.",
         body="This is a test email automatically generated by COS-alerter.",
+        dedup_key="testkey",
     )

k8s-local-test/deploy.yaml

@@ -0,0 +1,47 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: cos-alerter
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: cos-alerter
+  template:
+    metadata:
+      labels:
+        app: cos-alerter
+    spec:
+      containers:
+      - name: cos-alerter
+        image: <registry-ip>/<image-tag>
+        imagePullPolicy: Always
+        ports:
+        - containerPort: 8080
+        volumeMounts:
+        - name: config
+          mountPath: /etc/cos-alerter.yaml
+          subPath: cos-alerter.yaml
+      volumes:
+      - name: config
+        configMap:
+          name: cos-alerter-configmap
+
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: cos-alerter-configmap
+data:
+  cos-alerter.yaml: |
+    watch:
+      down_interval: "5m"
+      wait_for_first_connection: true
+      clients: {}
+    notify:
+      destinations: []
+      repeat_interval: "1h"
+    log_level: "info"
+    web_listen_addr: "0.0.0.0:8080"
+
+

k8s-local-test/svc.yaml

@@ -0,0 +1,13 @@
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: cos-alerter-service
+spec:
+  selector:
+    app: cos-alerter
+  ports:
+    - protocol: TCP
+      port: 8080
+      targetPort: 8080
+  type: ClusterIP

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "cos-alerter"
-version = "0.8.0"
+version = "0.9.0"
 authors = [
   { name="Dylan Stephano-Shachter", email="dylan.stephano-shachter@canonical.com" }
 ]
@@ -28,6 +28,7 @@ dependencies = [
   "timeago~=1.0",
   "waitress~=2.1",
   "xdg-base-dirs~=6.0.1",
+  "pdpyras~=5.2.0"
 ]
 
 [project.urls]

rockcraft.yaml

@@ -1,7 +1,7 @@
 name: cos-alerter
 summary: A liveness checker for self-monitoring.
 description: Receive regular pings from the cos stack and alert when they stop.
-version: "0.8.0"
+version: "0.9.0"
 base: ubuntu@22.04
 license: Apache-2.0
 platforms:

snap/snapcraft.yaml

@@ -1,5 +1,5 @@
 name: cos-alerter
-version: '0.8.0'
+version: '0.9.0'
 summary: A watchdog alerting on alertmanager notification failures.
 license: Apache-2.0
 contact: simon.aronsson@canonical.com