status.cocoapods.org

[orta]

A simple appfog/dotcoud app that receives pings from some cocoapods components and has a js page that does a look at whether things are up or not.

Requires:

• adding a DB for storing the last

externals

• adding a status command to search
• adding a status command to trunk
• adding a status command to cocoadocs

internals

• receives a ping for when cocoadocs last updated
• receives a ping for when search last updated
• receives a ping for when trunk last updated

web page

• JS check that heroku is up ( trunk / search )
• JS check that AWS is up ( cocoadocs files )
• JS check that github pages are up ( blog / cocoadocs.org / guides )

[orta]

Prototype it in https://app.ducksboard.com/ and see if that will deal with being the server.

[orta]

https://www.statuspage.io offered us a free account

[kylef]

I've just had a look around with StatusPage and it looks like a nice and flexible system. However we will also need a secondary "provider" to tie into StatusPage to provide metrics.

They support the following:

• Pingdom
• Librato
• New Relic - Have a free tier with low data retention (not sure what that means for us, I think that's fine since status page are retaining the data).
• TempoDB
• Datadog (they have a free tier for up to 5 hosts)

We can use "subscribers" to give us campfire notifications when something is down.

There is an API, however I can't determine if it let's us view a public status without authentication for use in the command line status. I've tweeted them and will update when I find out.

Custom metrics

We can also manually add metrics from code (example in Ruby). This means that trunk/search/cocoadocs can actually send data in such as the following metrics:

• CocoaDocs: Docset generations
• Trunk imports
• Search: Count of indexed pod's

Here are the rules for metrics:

• At minimum, one data point must be submitted every 5 minutes. If there is a lag in data, the gap will be reflected on the charts for the "Day" view.
• Each data point is cast to its nearest 30s interval, giving us a maximum of 10 data points per 5 minute period. Submitting multiple data points near each other will result in the last data point being the only one stored.
• Data can, and should, be backfilled up to 28 days in the past.

Example code in Ruby for sending a metric

require 'rubygems'
require 'httparty'

# the following 4 are the actual values that pertain to your account and this specific metric
api_key = 'REDCATED'
page_id = 'REDCATED'
metric_id = 'REDCATED'
api_base = 'https://api.statuspage.io/v1'

# need at least 1 data point for every 5 minutes
# submit random data for the whole day
total_points = (60 / 5 * 24)
total_points.times.each do |i|
  dhash = {
    :timestamp => Time.now.to_i - i * 5 * 60,
    :value => rand(100)
  }

  HTTParty.post(
    "#{api_base}/pages/#{page_id}/metrics/#{metric_id}/data.json", 
    :headers => {
      'Authorization' => "OAuth #{api_key}",
    },
    :body => {
      :data => dhash
    }
  )

  puts "Submitted point #{i+1} of #{total_points}"
end

Documentation: http://doers.statuspage.io/api/v1/metrics/

[kylef]

"Unofficial" way to view status page is the /index.json. This could be something we can pull from the pod command. Example: http://status.travis-ci.com/index.json

There sure is. adding /index.json to the end of any status page is our "unofficial" way right now but will give component status
If you're a customer, we provide an embedded javascript file you can use to show your status in a web app

https://twitter.com/statuspageio/status/436146802132537344
https://twitter.com/statuspageio/status/436147190135001089

[floere]

@kylef Can they also pull a metric?

[kylef]

@floere I can't find any API/method for setting that up, I think it's push only.

[orta]

This now exists 👍