AlertSnitch
Captures Prometheus AlertManager alerts and writes them in a MySQL or
Postgres database for future examination.
Because given a noisy enough alerting environment, offline querying
capabilities of triggered alerts is extremely valuable.
How does it work
- You stand up one of these however you like (multi-arch Docker images provided)
- You setup AlertManager to point at it and propagate your alerts in.
- Every alert that gets triggered reaches your database.
- Profit.
graph TD
A[alertmanager] -->|POST|B(AlertSnitch)
B --> |Save|C(MySQL/PG Database)
C -.-|Graph|G[Grafana]
C -.-|Query|D[MySQL/PG Client]
style B fill:#f9f,stroke:#333,stroke-width:1px
style C fill:#00A0A0,stroke:#333,stroke-width:1px
style D fill:#00C000
style G fill:#00C000
Local install
Simply install to your $GOPATH using your GO tools
$ go get github.com/daangn/db.AlertSnitch`
Requirements
To run AlertSnitch requires a MySQL or Postgres database to write to.
The database must be initialized with AlertSnitch model.
AlertSnitch will not become online until the model is up to date with the
expected one. Bootstrapping scripts are provided in the [scripts][./script.d]
folder.
Configuration
MySQL
For specifics about how to set up the MySQL DSN refer to Go MySQL client driver
This is a sample of a DSN that would connect to the local host over a Unix socket
export ALERTSNITCH_BACKEND="mysql"
export ALERTSNITCH_DSN="${MYSQL_USER}:${MYSQL_PASSWORD}@/${MYSQL_DATABASE}"
Postgres
export ALERTSNITCH_BACKEND="postgres"
export ALERTSNITCH_DSN="sslmode=disable user=${PGUSER} password=${PGPASSWORD} host=${PGHOST} database=${PGDATABASE}"
How to run
Running with Docker
Run using docker in this very registry, for ex.
$ docker run --rm \
-p 9567:9567 \
-e ALERTSNITCH_DSN \
-e ALERTSNITCH_BACKEND \
registry.github.com/daangn/db.AlertSnitch
Running Manually
- Open a terminal and run the following
- Copy the AlertSnitch binary from your $GOPATH to
/usr/local/bin
with sudo cp ~/go/bin/alertsnitch /usr/local/bin
- Now run AlertSnitch as with just
alertsnitch
- To just see the alerts that are being received, use the null backend with
ALERTSNITCH_BACKEND=null
Setting up in AlertManager
Once AlertSnitch is up and running, configure the Prometheus Alert Manager to
forward every alert to it on the /webhooks
path.
---
receivers:
- name: alertsnitch
webhook_configs:
- url: http://<alertsnitch-host-or-ip>:9567/webhook
Then add the route
# We want to send all alerts to alertsnitch and then continue to the
# appropiate handler.
route:
routes:
- receiver: alertsnitch
continue: true
Command line arguments
- -database-backend sets the database backend to connect to, supported are
mysql
, postgres
and null
- -debug dumps the received WebHook payloads to the log so you can understand what is going on
- -listen.address string address in which to listen for HTTP requests (default ":9567")
- -version prints the version and exit
Environment variables
- ALERTSNITCH_DSN required database connection query string
- ALERTSNITCH_ADDR same as -listen.address
- ALERTSNITCH_BACKEND same as -database-backend
Readiness probe
AlertSnitch offers a /-/ready
endpoint which will return 200 if the
application is ready to accept WebHook posts.
During startup AlertSnitch will probe the MySQL database and the database
model version. If everything is as expected it will set itself as ready.
In case of failure it will return a 500 and will write the error in the
response payload.
Liveliness probe
AlertSnitch offers a /-/health
endpoint which will return 200 as long as
the MySQL/Postgres database is reachable.
In case of error it will return a 500 and will write the error in the
response payload.
Metrics
AlertSnitch provides Prometheus metrics on /metrics
as per Prometheus
convention.
Security
There is no offering of security of any kind. AlertSnitch is not ment to be
exposed to the internet but to be executed in an internal network reachable
by the alert manager.
Grafana Compatibility
AlertSnitch writes alerts in such a way that they can be explored using
Grafana's MySQL/Postgres Data Source plugin. Refer to Grafana documentation
for further instructions.
Testing locally
We provide a couple of Makefile tasks to make it easy to run integration tests
locally, to get a full coverage sample run:
make bootstrap_local_testing
make integration
go tool cover -html=coverage.out
make teardown_local_testing