Garden Insights provides analytics for build, test, deploy and tasks executed by garden. The main metrics used are timings and outcomes over a 30 day time window. In order to make the data more actionable, we include a trend indicator and detailed view with timeseries for each action. For example, it might help to identify that the mean build time of a service has increased by 40% or the failures of the integration test suite has doubled compared to the previous 30 days.
Insights Quick Intro
The insights data is scoped by project and you find it by selecting a project and then clicking "Insights" in the side-navigation.
Insights Project Page
This is the main insights page, showing a table with summary statistics and trends by action type, name and environment. The table is sortable by each column and as shown in the screenshot below, it is possible to filter by environment by using a drop-down.
Insights Selectable Environment
When clicking on an action you'll be brought to a detail page containing the summary data over time. The first graph shows the average execution time per day for the selected action. In the second graph, the number of outcomes per category (success or failure) are stacked per day. Hovering over each day shows the exact daily numbers.
Below the charts, you'll find the full list of action for the interval displayed. Clicking on an action, will bring you to the Command page corresponding to when the action was run.
Insights Detail Graphs
You can display all the action for a shorter interval selecting the dates to analyze in the charts.
Insights Detail Interval Selection
Details on Data
The data is aggregated over a window of the previous 30 days. Trend percentages and indicators are compared against the previous 30 days. The data is calculated using a batch job run on a daily basis and the date of the last successful run is displayed up to the right.
The duration aggregates contain all actions independent of their outcome. This means that the mean and p95 includes failed or cached actions. The reasoning behind this is that the durations indicate how long you would need to wait on average for the action to complete. Below we describe all the columns in the table.
Action, the executed action including build, test, deploy and task.
Name, the name of the action target.
Environment, the environment the action was executed in.
Duration (mean), The average duration of the executed action, it includes cached actions.
Duration (p95), The 95th percentile duration of the executed action.
Duration Trend, The duration trend, calculated against the mean over the previous time window.
Total, The total number of executions for the given period.
Failure Rate, The ratio of errors over total executions over the last period.
Failure Trend, The failure trend, calculated against the failure rate over the previous time window.