Dashboards let you visualize any metric received by Kamino, including metrics automatically collected by the instrumentation and metrics that you manually record. You can create as many dashboards as you want, with as many charts as you like.
Creating a Dashboard
To create a Dashboard, go to the Dashboards page by clicking the link on the sidebar. Once on the Dashboards page click Create:
Selecting a Metric
There are five basic pieces of information that you must provide in order to create a chart in the dashboards:
- Service: Select the service from which metrics will be pulled.
- Metric: The full metric name you want to include.
- Filter Tags (optional): If you would like to display only time series that match a specific tag you can include them here.
- Merge Series (optional): Allows you to decide whether all time series matching the query should be shown independently (disabled) or if they should be merged into a single result (enabled).
- Group By (optional): If Merge Series is enabled, the Group By setting allows you break down the results using the specified tags.
Let's look at an example. In the configuration bellow we are selecting the metric "span.processing-time" from the "event-ticketing" service. Then we making sure that only time series with the tag "error=false" are included and we group them by host and operation.
You can see how the resulting line chart has one line per each unique combination of host and operation tags.
At a minimum, you can provide a Custom title, measurement unit (like Latency, Information, etc) and a Chart Width. Depending on the metric you select, different Chart Type options will appear.
Line charts use a Value Function to determine what value to draw in the charts. Minimum, Maximum and Percentiles are some of the available options.
Percentile Areas Chart
This chart displays areas covering the common percentiles of interest:
- From the minimum and up to 50th percentile.
- From the 50th percentile to the 90th percentile.
- From the 90th percentile to the 95th percentile.
- From the 95th percentile to the 99.9th percentile.
Stacked Bars Chart
Works similarly to the Lines Chart, but all values are stacked on top of each other rather than shown at the same scale. This is specially useful when looking at global throughput or counters.
Displays a histogram with all values from the selected metric.
You will notice a small underline bellow the X axis which indicates that there are values on the bucket covered by that area. This helps you acknowledge the existence of data in cases like the above where the number of recordings on the leftmost buckets are so big that the outliers' bars are practically invisible.
This chart can be considered an alternative representation of the Value Distribution chart. Here, the X axis represents the full spectrum of percentiles up to six nines (99.9999%) and the value of each one of those percentiles.
One of the biggest benefits of this chart type is that it draws attention to the high percentile values, which is where the outliers and unhappy users are.
You can think of heatmaps as a "histograms over time" representation. The distribution of values is analyzed for each time bucket and a darker shade is used for the color where most of the recordings fell whilst a brighter shade is used where the least recordings are.
The Areas chart draw an area that goes from the "Low Boundary" value up yo the "High Boundary" value. This is useful when trying to compare the operating range of a given metric across several time series. Typically you would be using it with the Minimum and Maximum or Percentiles functions.
Editing and Removing Charts
To change any chart configuration you must go into Edit mode but clicking on the Edit button on the upper-right corner of the Dashboard Page, then you will how all charts show a small ribbon with the edit or delete buttons:
You can the edit and/or remove charts. Remember to click Save once you are done!