ben/webhook-matrix-notifier
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

Add better output for prometheus alerts

Browse source 
and add testrequest JSON file.
This commit is contained in:
Ben 2020-09-27 17:55:20 +02:00
parent c3766767ba
commit 5c7571f952
Signed by: ben
GPG key ID: 0F54A7ED232D3319
4 changed files with 123 additions and 51 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
README.md
View file

@ -7,9 +7,8 @@ behind a reverse-proxy with HTTPS.
# Testing the Hook locally
- Start the webserver locally by `env FLASK_APP=wmn.py flask run`
- Or have your IDE do it for you
- Send a POST request using curl `curl -i -X POST "localhost:5000/matrix?channel=%21yhEUnvhAZZFKRStdXb%3Amatrix.org" -H "X-Gitlab-Event: Push Hook" -H "X-Gitlab-Token: ..." -H "Content-Type: application/json" --data-binary @./testrequest.json`
- Send a POST request using curl `curl -i -X POST "localhost:5000/matrix?channel=%21yhEUnvhAZZFKRStdXb%3Amatrix.org" -H "X-Gitlab-Event: Push Hook" -H "X-Gitlab-Token: ..." -H "Content-Type: application/json" --data-binary @./testrequest_gitlab.json`
- The part after `channel=` is the room ID which can retrieved from Matrix channels you are part of
- `%21` escapes ! in URI
- `%3A` escapes : in URI
- The `X-Gitlab-Token` must correspond to the one provided in `config.yaml`

0
testrequest.json → testrequest_gitlab.json
View file

37
testrequest_prometheus.json Normal file
View file

@ -0,0 +1,37 @@
{
"receiver": "webhook",
"status": "firing",
"alerts": [
{
"status": "firing",
"labels": {
"alertname": "Test",
"dc": "eu-west-1",
"instance": "prometheus.example.com:9090",
"job": "prometheus24"
},
"annotations": {
"description": "some description"
},
"startsAt": "2018-08-03T09:52:26.739266876+02:00",
"endsAt": "0001-01-01T00:00:00Z",
"generatorURL": "http://prometheus.example.com:9090/graph?g0.expr=go_memstats_alloc_bytes+%3E+0\u0026g0.tab=1"
}
],
"groupLabels": {
"alertname": "Test",
"job": "prometheus24"
},
"commonLabels": {
"alertname": "Test",
"dc": "eu-west-1",
"instance": "prometheus.example.com:9090",
"job": "prometheus24"
},
"commonAnnotations": {
"description": "some description"
},
"externalURL": "http://prometheus.example.com:9093",
"version": "4",
"groupKey": "{}:{alertname=\"Test\", job=\"prometheus24\"}"
}

134
wmn.py
View file

@ -10,7 +10,8 @@ from flask import Flask, request, abort
from matrix_client.client import MatrixClient
from matrix_client.errors import MatrixRequestError
application = Flask(__name__)
app = Flask(__name__)
application = app
# Not going to care for specifics like the underscore.
# Generally match room alias or id [!#]anything:example.com with unicode support.
@ -63,6 +64,10 @@ def get_msg_type():
abort(400)
def color_format_html(color_hex: str, text: str):
return f'<font color="#{color_hex}">{text}</font>'
def iter_first_line(string: str):
return iter(map(str.rstrip, string.lstrip().splitlines(keepends=False)))
@ -203,6 +208,9 @@ def process_prometheus_request():
msgtype = get_msg_type()
room = get_a_room()
if not request.json:
abort(400)
# written for version 4 of the alertmanager webhook JSON
# https://prometheus.io/docs/alerting/configuration/#webhook_config
@ -210,10 +218,13 @@ def process_prometheus_request():
_status_colors = {"resolved": "34A91D", "firing": "EF2929"}
if text is None:
text = status
if status in _status_colors:
return f'<font color="#{_status_colors[status]}">{text}</font>'
else:
return text
return color_format_html(_status_colors.get(status, "FFFFFF"), text)
def color_severity_html(severity: str, text: typing.Optional[str] = None):
_severity_colors = {"warning": "EFAC29", "critical": "EF2929"}
if text is None:
text = severity
return color_format_html(_severity_colors.get(severity, "FFFFFF"), text)
def parse_promtime(date_string):
match = promtime_to_isotime_pattern.match(date_string)
@ -225,63 +236,88 @@ def process_prometheus_request():
grps[-1] = '+00:00'
return datetime.fromisoformat(''.join(grps))
def alert_title(status: str, alertname: str, generator_url: str):
if alertname:
alertname = " alert " + alertname
if status:
status_msg = status.upper() if status == "firing" else status.title()
title = status_msg + alertname
html_title = color_status_html(status, title)
elif alertname:
title = alertname
html_title = title
else:
title = ""
html_title = title
if title:
title = f"*{title}*"
if generator_url:
title = f"{title} {generator_url}"
if html_title:
html_title = f"<strong>{html_title}</strong>"
if generator_url:
html_title = f'<a href="{generator_url}">{html_title}</a>'
return title, html_title
def extract_alert_message(alert: typing.Dict[str, typing.Any]) -> typing.Tuple[str, str]:
"""Takes the alert object and returns (text, html) as a string tuple."""
alert_status = alert.get("status", "None")
alert_labels = str(alert.get("labels", None))
alert_annotations = str(alert.get("annotations", None))
alert_start = alert.get("startsAt", None)
alert_end = alert.get("endsAt", None)
labels = alert.get("labels", {})
severity = labels.get("severity", "")
annotations = alert.get("annotations", {})
description = annotations.get("description", "")
if not description:
description = annotations.get("summary", "")
alert_daterange = []
if alert_start is not None and alert_end != '0001-01-01T00:00:00Z':
alert_start = parse_promtime(alert_start).strftime("%d. %b %y %H:%M %Z").rstrip()
alert_daterange.append(f'Started at {alert_start}')
if alert_end is not None and alert_end != '0001-01-01T00:00:00Z':
alert_end = parse_promtime(alert_end).strftime("%d. %b %y %H:%M %Z").rstrip()
alert_daterange.append(f'Ended at {alert_end}')
alert_daterange = "" if len(alert_daterange) == 0 else f'({", ".join(alert_daterange)})'
alert_generator_url = alert.get("generatorURL", "None")
if "startsAt" in alert and alert["startsAt"] != '0001-01-01T00:00:00Z':
alert_start = parse_promtime(alert["startsAt"]).strftime("%d. %b %y %H:%M %Z").rstrip()
alert_daterange.append(f'started at {alert_start}')
if "endsAt" in alert and alert["endsAt"] != '0001-01-01T00:00:00Z':
alert_end = parse_promtime(alert["endsAt"]).strftime("%d. %b %y %H:%M %Z").rstrip()
alert_daterange.append(f'ended at {alert_end}')
alert_daterange = ", ".join(alert_daterange)
return (
f'[{alert_status}] Labels: {alert_labels}, Annotations: {alert_annotations} - {alert_daterange} | Generator: {alert_generator_url}',
f'<strong>{color_status_html(alert_status)}</strong> Labels: {alert_labels}, Annotations: {alert_annotations} - {alert_daterange} | Generator: {alert_generator_url}',
title, html_title = alert_title(
status=alert.get("status", ""),
alertname=labels.get("alertname", ""),
generator_url=alert.get("generatorURL", "")
)
if severity:
html_severity = f"Severity: {color_severity_html(severity)}"
severity = severity.upper() if severity == 'critical' else severity.title()
severity = f"Severity: {severity}"
else:
html_severity = ""
def extract_prometheus_message() -> typing.Tuple[str, str]:
"""Dissects the request's JSON and returns (text, html) as a string tuple."""
group_key = request.json.get("groupKey", "None")
status = request.json.get("status", "None")
receiver = request.json.get("receiver", "None")
group_labels = str(request.json.get("groupLabels", None))
common_labels = str(request.json.get("commonLabels", None))
common_annotations = str(request.json.get("commonAnnotations", None))
ext_url = request.json.get("externalURL", "None")
alerts = request.json.get("alerts", []) # type: typing.List[typing.Dict[str, typing.Any]]
text_alerts, html_alerts = zip(*map(extract_alert_message, alerts))
text_alerts = "\n" + "\n".join((f"- {msg}" for msg in text_alerts))
html_alerts = "<br>\n<ul>\n" + "\n".join((f" <li>{msg}</li>" for msg in html_alerts)) + "\n</ul>"
html_parts = [html_title, html_severity, description, alert_daterange]
html_message = "</p>\n<p>".join(filter(bool, html_parts))
html_message = f"<p>{html_message}</p>" if html_message else ""
return (
f'*{status.title()} alert for group {group_key}*\n Receiver: {receiver}\n Labels: {group_labels} | {common_labels}\n Annotations: {common_annotations}\n External URL: {ext_url}\nAlerts:{text_alerts}',
f'<strong>{color_status_html(status, f"{status.title()} alert for group {group_key}")}</strong><br>\n <em>Receiver:</em> {receiver}<br>\n <em>Labels:</em> {group_labels} | {common_labels}<br>\n <em>Annotations:</em> {common_annotations}<br>\n <em>External URL:</em> {ext_url}<br>\n<em>Alerts:</em>{html_alerts}',
" \n".join(filter(bool, [title, severity, description, alert_daterange])),
html_message
)
try:
html, body = extract_prometheus_message()
except (LookupError, ValueError, TypeError):
print("Error parsing JSON and forming message:", file=sys.stderr)
traceback.print_exc()
print(file=sys.stderr, flush=True)
return "Error parsing JSON and forming message", 500
try:
client = MatrixClient(cfg["matrix"]["server"])
client.login(username=cfg["matrix"]["username"], password=cfg["matrix"]["password"])
room = client.join_room(room_id_or_alias=room)
room.send_html(html=html, body=body, msgtype=msgtype)
try:
for body, html in map(extract_alert_message, request.json.get("alerts", [])):
if html and body:
room.send_html(html=html, body=body, msgtype=msgtype)
elif body:
room.send_text(body)
except (LookupError, ValueError, TypeError):
room.send_text("Error parsing data in prometheus request")
print("Error parsing JSON and forming message:", file=sys.stderr)
traceback.print_exc()
print(file=sys.stderr, flush=True)
return "Error parsing JSON and forming message", 500
except MatrixRequestError as e:
return matrix_error(e)
@ -289,7 +325,7 @@ def process_prometheus_request():
return "", 204
@application.route('/matrix', methods=("POST",))
@app.route('/matrix', methods=("POST",))
def notify():
if 'X-Gitlab-Token' in request.headers:
return process_gitlab_request()