@SamFritz @ianmilligan1 let me know what you think of this:

That should be all the bullet points above, plus jobs run over time.

I'm using Chartkick.

• Users is an area chart (by day - I'd probably change this to month)
• Jobs is a multiple series* (by month)
• The rest are a scatter plot

I'm at York tomorrow, so no rush.

(Oh, you check out that Chartkick page, you'll see the option to download the charts as images. I set that up for each of them, so y'all can use any of those in presentations in the future 😉 )

This is looking great @ruebot! Really like how the images are easily downloadable for additional use :)

Just wondering for graphs that have an access labeled 2018.0; 2018.1; 2018.2, etc, does the decimal point indicate the day of the year. e.g 2018.3 = January 3, 2018?

Throughput graphs updated:

• x-axis: id (just an auto-incrementing number in the job dashboard table)
• y-axis: job time in minutes
• First 4 throughput graphs are: scatter, area, line, column

Let me know which of those 4 you like the best, then we can work on axis labels.

"jobs" should have smoothing; in fact, none of the graphs should.

@lintool should or shouldn't?

Sorry - shouldn't

Curves/smoothing removed.

Also, "throughput" probably isn't the best word to use here for all the "throughput" charges. Those are all just showing how long each job was.

This is the method that basically powers all of them:

  def job_times(queue_name)
    jt = Dashboard.where(queue: queue_name)
                  .where('end_time is not null')
                  .pluck(:end_time, :start_time, :id)

    jt.map { |k, v, i|
      Hash[i, TimeDifference.between(k, v).in_minutes]
    }.inject(:merge)
  end

These look great, @ruebot – esp. without curves/smoothing. At a glance it gives a fantastic overview of where we're at with our different processes and elements, which is what we like dashboards to do.

As for throughput graph style, I would have a preference for area or line. Scatter requires more effort to read and column is a bit boxy.

Ok, I'm going to commit, and push this up to production here in a moment. Then we can do further feedback from there as y'all will be able to interact with it will production data.

Updated! Check it out, and drop feedback in here.

Caught a couple things that might be helpful:

• Title of stats page isn't consistent (updated locally)
• Added id column to jobs table, so you can see which job it is from the graph.

Maybe "User Registrations" graph should go by month, not day.

• Rename all "throughput" graphs to "<queue_name> Job Length"
• Add actual throughput graphs for each queue; job time over collection size (GB)

These look fantastic, @ruebot - as you'll see from the Slack I was nerding out and excitedly screenshooting.

A few thoughts:

Maybe "User Registrations" graph should go by month, not day.

That's probably most useful. When we have clusters of sign ups, we can notice it on the jobs dashboard, whereas the "user registrations" I think could give us good long-term data?

Rename all "throughput" graphs to "<queue_name> Job Length"

Good call – that's probably more understandable at a glance.

Rename all "throughput" graphs to "<queue_name> Job Length"

👍 that'd be really useful!

New charts look great! Tooltips work really well, the user registrations per month are very handy and good to see our long-term growth, and the throughput and job lengths are legible and really useful to me.

Resolved with:

• 00d21d1
• 7143dd7
• 542e61e
• bc6fef7