I’m pleased to announce Juju Metrics, new in Juju 2.0! Knowing an application's configuration isn’t enough to effectively operate and manage it. Consider that a well-designed application will have as few configurable parameters as possible. As an operator, you may find yourself wanting to know more about the resources that an application in your model consumes and provides -- resources such as:
- Storage GiB used - Number of user accounts - Number of recently active users - Active database connections Juju Metrics complete the operational picture with application observability; by modeling, sampling and collecting measurements of resources such as these. Juju collects application metrics at a cadence appropriate for taking a model-level assessment of application utilization and capacity planning. Charm authors and communities can now collaborate on figuring out the critical metrics for service assurance and ops, and share that through the charms. There are many instrumentation and time-series data collection solutions supporting devops. Juju’s metrics complement these fine-grained, lower-level data sources with a model-level overview -- possibly a starting point for deeper analysis. At a glance, operators can pull the most recent measurements across the entire model: $ juju metrics --all UNIT TIMESTAMP METRIC VALUE auth-sso/0 2016-09-19T22:14:29Z users 28 auth-sso/0 2016-09-19T22:14:31Z tokens 6 ceph-mon/0 2016-09-19T22:15:36Z gb-usage 5.2711902345 webapp/0 2016-09-19T22:17:57Z requests 11903 webapp/1 2016-09-19T22:17:52Z requests 13719 View measurements for specific units or applications: $ juju metrics webapp/0 UNIT TIMESTAMP METRIC VALUE webapp/0 2016-09-19T22:17:57Z requests 13719 $ juju metrics sso-auth auth-sso/0 2016-09-19T22:14:29Z users 28 auth-sso/0 2016-09-19T22:14:31Z tokens 6 How about adding metrics to a charm? With the reactive framework, it’s quick and easy! 1. Add layer:metrics to your charm’s layer.yaml 2. Declare measurements in your charm’s metrics.yaml metrics.yaml declares each metric’s type and the command line that measures it. For example, the hypothetical auth-sso charm in the example above might declare metrics: metrics: users: type: gauge description: Number of users command: scripts/count_users.py tokens: type: gauge description: Number of active tokens command: scripts/count_tokens.py The command lines given in command: attributes above simply need to write the current gauge value -- a positive decimal number -- to standard output. Juju 2.0 will initially support type: gauge metrics for the operational use cases shown above; others such as type: absolute are experimental, and will be better supported in Juju 2.1. I’d like to encourage you to kick the tires on this new feature, and let’s start measuring things in our charms! For more information, here are links to documentation: Metrics, User Guide: https://jujucharms.com/docs/stable/charms-metrics Metrics, Developer Guide: https://jujucharms.com/docs/stable/developer-metrics -Casey
-- Juju mailing list Juju@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/juju
