Weekly Development Summary

[Ian Booth]

It's been a busy week for Juju this week.

The big news is that after 4 betas and a few release candidates, we released
Juju 2.2.0 into the wild [1].

There's a number of great new features, including support for Oracle cloud and
much improved support for vSphere, and a nicer Azure login experience.

One aspect of our work for this release worth calling out is the focus on
improving Juju's scalability - both in terms of running many small/medium
models, as well as a few much larger models with thousands of unit agents. We're
not done yet, and a few improvements missed the deadline, so we're looking to
release a 2.2.1 version end of next week.

Areas where we've made changes so far include:
- throttling controller connections under load
- much improved metrics to allow us to identify scalability issues
- knobs to tune log and status history pruning
- splitting model into separate collections

With the new metrics, we used Prometheus, Telegraf, and Grafana to create a nice
dashboard to monitor and visualise system behaviour. There's a great blog post
by Rick Harding on this general topic [2]. Expect more communications soon
describing the new metrics and dashboards we used in the real world to do this
work. The best bit - using Juju itself makes visualising a running Juju
controller's metrics super easy.

[1] https://jujucharms.com/docs/stable/whats-new
[2] http://mitechie.com/blog/2017/3/20/operations-in-production

Quick links:
  Work pending: https://github.com/juju/juju/pulls
  Recent commits: https://github.com/juju/juju/commits/develop