Skip to content

Latest commit

 

History

History
293 lines (194 loc) · 11.1 KB

settings.md

File metadata and controls

293 lines (194 loc) · 11.1 KB
Raw

Settings

The settings.yml file is divided into three main sections

transports:
    ...

recipients:
    ...

watchers:
    ....

Transports

Transports are used to send alerts when watcher states change.

SMTP Transport

You can send email using any SMTP server. AWD? uses simple user/password SMTP authenticaiton, and has been confirmed working with Amazon SES and MailGun. For example, to set up Mailgun as a transport, use

transports:
    smtp:
        server : smtp.mailgun.org
        port : 587
        secure : false
        user : postmaster@<your-mailgun-account>.mailgun.org
        pass: your-mailgin-password
        from : [email protected]

With that out of the way, it should be said that SMTP is really simple to set up if you're doing everything exactly right, and difficult to diagnose the instant something goes wrong. If you're struggling with settings to work, first test them independently with a simple SMTP client like swaks:

    sudo apt install swaks -y

    ./swaks --auth \
        --server smtp.some-service.com \
        --au postmaster@YOUR_DOMAIN_NAME \
        --ap your-password \
        --to [email protected] \
        --h-Subject: "Hello world" \
        --body 'Testing send you some mail'

If that works, AWD? has its own STMP test that sends your a mail directly. Use it with

    curl http://yourAWDserver/diagnostics/transport/smtp/<recipient-node>

where recipient-node is the recipient node in AWD's config with an smtp:email-address child property on it.

Slack Transport

Alerts can be sent to a Slack user or channel using a custom app that acts as a bot - check out apps dashboard.

  • Your Slack app needs four scopes, channels:read, groups:read, mpim:read, im:read, chat:write. To post to channels you also need to manually add your app to the channel in question, via the Slack app.

  • Secret is also called "Signing Secret" is specific to your Slack app.

  • Token normally starts with xox and is specific to your app's integration into your workspace. To find the token, open "add features and functionality" then click on "permissions", then look for "Bot User OAuth Token".

      transports:
      slack:
          secret: ...............
          token: xoxb-...........

Recipients

Recipients are people or systems that receive alerts.

SMTP

To receive emails add smtp to a recipient and add email address to it

recipients:
    BobMcName:
        smtp: [email protected]

Slack

To receive Slack alerts add slack to a recipient (either a person or something imaginary for channel alerts) and define either a Slack userid or channelid.

recipients:

    BobMcName:
        slack: <BobMcName's slack userId>

    foo:
        slack: <slack channelId>

To get a user id, click on the user profile in the Slack desktop client, then look under "More". To find a channel id, open Slack in a browser and click on the channel you want to post to - the channel id is the second id in browser address bar https://app.slack.com/client/workspace-id/channel-id/user_profile/your-own-user-id

Watchers

Watchers let you test something at regular intervals. The following built-in tests are available.

HTTP 200

The simplest (and default) watcher is the HTTP check. It queries a host and fails if it doesn't get an HTTP code 200 back. You can specify another HTTP code if you expect something other than 200 (such as a 403 if the host prompts you to login).

watchers:
    mytest:
        host: example.com

        # optional code if you're expecting something other than 2xx
        code : 403

Port open

Test if a port at the given host is open. Works on TCP only.

watchers:
    port:
        test: net.portOpen
        host: 192.168.1.126
        port: 8006

Jenkins job status

Test if a jenkins job is passing. Requires a Jenkins server URL and job name.

  • Host can be any URL that gives access to the server, this is often with built-in credentials.

  • Job name can be the human-friendly version, we'll make it URL-safe, so copy this directly from your Jenkins UI if you want.

  • By default, a test passes on success only, all other outcomes will be read as failure. You can override this with status value, which can be a comma-separated string consisting of success, aborted and/or failure (these strings are taken directly from the Jenkins API).

      watchers:
      my_jenkins_job:
          test : jenkins.buildSuccess
          host: <USER>:<PASSWORD>@example.com
          job: My Jenkins Job

          # Optional.
          status: success,aborted

Docker container up

If your have the Docker HTTP API enabled, you can query it to test if a container is up on a given docker host.

watchers:
    my-container-test:
        test: docker.containerIsUp
        host: example.com
        container: myContainer # container name

        # Optional. Port to query, the default is 2375.
        port: 2375

System.d service running

You can test if a system.d service is running. You will need SSH access to the machine running the service. Password can be templated in via an env var (see advanced settings).

watchers:
    my-service-test:
        test: systemd.servicerunning
        host: example.com
        user: myuser
        password: mypassword

        # name of the system.d service to query
        service: docker

Disk use

You can check if a disk is runnnig out of space (Linux only). You will need SSH access to the machine running the service. Password can be templated in via an env var (see advanced settings).

watchers:
    my-disk-use-test:
        test: resource.linux.diskUse
        host: example.com
        user: myuser
        password: mypassword
        path: /some/path/on/host

        # max % use allowed
        threshold: 50

Ping

You can ping a host.

  • Timeout is in seconds, and is optional.

      watchers:
      my-ping-test:
          test: net.ping
          host: example.com

          # optional
          timeout: 10

ZFS Pool Status

You can query the status of a ZFS pool on a host. This test ensures that the pool's state is online. You will need SSH access to the machine running the service. Password can be templated in via an env var (see advanced settings).

watchers:
    is-my-data-gone:
        test: filesystem.zfs.zpoolStatus
        host: example.com
        user: myuser
        password: mypassword

        # name of zfs pool to check
        pool: mypoolname

Date In File

You can rig up a simple "dead man's switch" by writing a date to a file at the end of some process. As long as the date in the file is current, the test will pass. The date written must be in Javascript-parsable ISO format. You can generate a file in a Linux terminal with the command

echo $(date --iso-8601=seconds) >> /path/to/datefile

You will need SSH access to the machine running the service. Password can be templated in via an env var (see advanced settings).

watchers:
    is-my-process-alive:
        test: general.dateInFile
        host: example.com
        user: myuser
        password: mypassword

        # path on remote machine date is written to
        path: /path/to/datefile

        # Maximum allowed age of date in file. Can be any digit followed by S, M, H or D (seconds, minutes, hours or days, case not important)
        range: 24H

Default watcher behaviour

  • Watchers have a default interval of 1 minute. You can override any watcher's default with any valid cronmask (masks must be in quotes, this is a YML quirk).

  • The YML node is the default name of a watcher. You can provide a display name using the optional name field.

  • Watchers, like most other objects in AreWeDown? have an optional enabled field that defaults to true. Set this to false to disable the watcher.

  • To alert specific people about a watcher failure, use recipients - this is an optional, comma-separated list of names defined under the top-level recipients section. If left empty, all recipients will receive alerts for that watcher.

      watchers:
      mytest:
          
          # Give it a fancier name with spaces and things
          name: my "fancy" test name!

          # Send alerts to these people only. Sane spacing optional (Python trigger warning)
          recipients: BobMcName,someOtherPerson,      YetAnotherPerson            

          # Run it on a Tuesday only, because.
          interval: '0 0 * * TUE'
                          
          # Actually, don't run it at all.
          enabled: false

Custom tests

You can write your own tests in any shell script supported by your host system. See custom tests for more.

Advanced Settings

Defaults

The following overridable default settings live in the root-level of settings.yml.

# default dashboard title 
header: Are We Down?

# defines where logs are written. Default value is relative to application startup path. If you want to 
# write to /var/log/arewedown for example, change this
logs: ./logs

# Port HTTP server run ons   
port: 3000

# Interval for dashboard refreshes (milliseconds)
dashboardRefreshInterval: 5000,

# Interval for dashboard timeout (milliseconds)
dashboardLoadTimeout: 5000,

# Allows app to be restarted from user interface
UIRestart: false

# Amount of data logged. Can be error, warn, info
logLevel: warn

# internal work cleans up/maintains self. needs to run once a day only. Must be wrapped in quotes.
internalWorkerTimer : '0 0 * * *'

# In days. AreWeDown? can clean up its own log data to prevent your disk from flooding.
logRetention: 365

Secrets

If you do not want to store sensitive information like passwords in the settings.yml file, you can pass these as environment variables from your host system, and bind them anywhere in settings with the "{{env.___}}" template pattern

anyProperty: "{{env.MY_SENSITIVE_INFO}}"

If you defined an environment variable MY_SENSITIVE_INFO=1234, the anyProperty setting have the value 1234 in memory. If a templated environment variable is not found, AreWeDown? will fail to start.

Logs

AreWeDown? logs, a lot. It writes its own logs to the /etc/arewedown/logs/<DATE>.log, and then for each watcher in /etc/arewedown/logs/<WATCHER>/logs/<DATE>.log. If AreWeDown? fails to start or exits abruptly, a good place to start looking is its logs.