Graphite Dashboard Guide

Contents

• Graphite Dashboard Guide

  • Creating a Graph

  • Choosing a Time Range

  • Graph Labeling and Styling

  • Creating a Graphite Dashboard

  • Saving and Loading Dashboards

  • Advanced Dashboard Configuration

NOTE: We have removed Graphite Composer from the HG application in favor of Grafana. If you require Graphite Composer for any reason, please reach out to us and we will activate it in your account.

In the following guide, we’ll walk you through creating a simple Graphite graph for a chosen metric and adding that to a dashboard that you can bookmark for quick access. If you haven’t sent any metrics to Hosted Graphite yet, check out our Sending Metrics Guide.

In this example, we’re using a metric called conc_users - the number of users logged into our imaginary application. We have been sending the metric value (such as the one below) once a minute to Hosted Graphite for the last day.

conc_users 1215

Creating a Graph

Let’s set up a graph for our conc_users metric. Log into Hosted Graphite, navigate to Dashboards, and select Graphite Composer.

When that opens, you’ll be presented with a blank composer dialog. On the tree on the left-hand side, expand the ‘Graphite’ folder. This will contain all the metrics sent using the API Key associated with your account. At the moment, there’s only one entry - our demo conc_user metric, so we click on that.

Choosing our conc_user metric

Tip: If you’re going to be sending a wide range of metrics to a Hosted Graphite account, this ‘Graphite’ folder could soon have an unmanageable number of entries. Using periods in your metric names will introduce a folder structure, allowing for logical grouping of your metrics. E.g. servers.mario.memory.free will result in Graphite\servers\mario\memory\free in the tree above. See the Graphite Project website for more info.

When we click on our conc_users metric, Graphite will render a graph of a recent selection that metric data:

A graph of the conc_user metric after clicking on the conc_user entry

Choosing a Time Range

The next thing we do is to choose how much of our data we’re going to sample. We have two options, via the following icons:

‘Select A Date Range‘ and ‘Select Recent Data‘ buttons

• Select A Date Range

  Choose a start date/time and an end date/time. This is useful for debugging a new metric, or for investigating an incident that occurred during a specific time window.

• Select Recent Data

  View a rolling window of the most recent metric data, up to a specified number of minutes/hours/days/weeks/months/years.

Tip - If you make changes but the graph doesn’t update automatically, use the Update Graph button to trigger a refresh of the graph.

‘Update Graph‘ Button

In this example, we want a view of the last 12 hours of the conc_users metric, so click Select Recent Data and enter the appropriate values.

Graph Labeling and Styling

There are many options to make your data more presentable. At the bottom of the composer window, you’ll find a group of menus. These contain various functions to let you apply data transformations, labeling, and styling to your graph:

Composer Menus

In this example, we’ll use only a few of the many available options. You can find more information about the menus in the graph menu guide. Let’s make the following adjustments to our graph:

• Graph Options -> Line Mode -> Connected LineMakes our graph a continuous line through our available metric points.

• Graph Options -> Graph TitleSets the title to Site Visitors.

• Graph Options -> Display -> Line ThicknessSet this value to 3 (pixels) for a slightly thicker line than the default.

• Graph Options -> X-Axis -> TimeZoneMetric timestamps are sent in UTC. The server (in this example) sending the metrics is in Dublin - Set this to Europe/Dublin to apply the appropriate offset to the X-axis timestamps.

• Graph Options -> Y-Axis -> LabelSet the vertical axis label to Users.

This results in the following:

A improved graph of the conc_user metric

Click on the floppy disk icon in the composer toolbar to save. You’ll be prompted for a name, enter Site Visitors, and click OK. Your graph is now saved and can be recalled at any time by browsing My Graphs in the left-hand window of the Composer.

Creating a Graphite Dashboard

The final step in this tutorial is to create a bookmarkable Graphite dashboard and add your graph for display. Having followed the tutorial above, we will have a new Site Visitors graph in our account. Navigate to Dashboards -> Graphite Composer -> Dashboard in your Hosted Graphite account.

When you first open the dashboard page, you’ll be presented with a blank dashboard. Go to Graphs -> New Graph -> From Saved Graph

Loading our saved Graph

In the dialog that opens, an entry called Site Visitors will appear. Choose it and then click Select. Your graph will appear in the dashboard. Let’s make our graph a little bigger. Click on Graphs -> Resize, and choose the Custom option. Enter in a width of 800 and a height of 500 (or whatever suits your display).

Tip You can further customize the graphs for this dashboard instance by clicking a graph once to focus it, and clicking once more to bring up a configuration menu (This is the same menu we used when creating the graph - See Graphite Graph Menu Reference for more info).

Our Final Dashboard!

Saving and Loading Dashboards

To save your finished dashboard, click Dashboard -> Save As and enter a name, e.g. MainSiteDash.

To get a bookmarkable URL (which will directly open this dashboard instance), click on Share in the menu. A dialog will appear (you may be asked to save again) containing a URL that you can copy.

Finally, you can see all dashboards that you own by clicking Dashboard -> Finder. A dialog will appear with a list of all the dashboards that you have saved.

Advanced Dashboard Configuration

Graph Data

In Creating A Dashboard, we created a simple graph with some basic styling. Graphite offers many more configuration options. You have the ability to show multiple metrics in the same graph, apply transformations to the metric data, or combine metrics before rendering. All of these advanced features are found via the Graph Data button:

Graph Data button

Each time a metric is selected (in the tree of metrics), an entry called a ‘target’ is added to the Graph Data dialog. For example, had we clicked on metric \servers\smithers\users_authed and then metric \servers\smithers\users_guest in the composer, the Graph Data dialog would contain two targets as follows (which would each result in a line on the graph):

Graph Data button

Wildcards

Sometimes we want to display groups of common metrics. Rather than manually adding each metric as a target, we can use wildcards to specify multiple metrics as one target. Using the Edit button in the Graph Data dialog, we can manually specify the target as in the following examples:

• To graph all the metrics under the smithers server:

  Set the target as servers.smithers.*

• To graph the users_authed metric on all servers:

  Set the target as servers.*.users_authed

Graph Performance

If you have many metrics stored with us, and your graphs are taking a long time to load, chances are that your wildcard use is the first place you should look. Try to avoid wildcards such as the following:

servers.*.*.cpu

In this case, every single metric matched by this wildcard will be queried, even if many of those metrics don’t have data for the selected time range. Multiple, but narrower targets in a graph may result in faster load times than one widely defined target.

In our example above, perhaps all we care about are our metrics regarding public-facing machines, but our wildcard is also including metrics from our production or staging machines. In this scenario, we replace the single target with two narrower targets, e.g.:

servers.production.*.cpu
servers.staging.*.cpu

Data Transformations

Now that we’ve defined which metrics we’re interested in, we can perform transformations on them. Back in the Graph Data dialog, when we select a target (or multiple targets), we can use the Apply Function menu to alter or combine our targets in some way. When a function is applied to a target, you’ll notice that the target’s text changes, eg

before: original_target_text

after:  functionName(original_target_text)   or
        functionName(original_target_text, optional_params..)

Eg, if we selected conc_users, and clicked Apply Function –> Transform –> Derivative, the target text would now say

The graph should automatically update to show the derivative of our conc_users metric. If not, click the refresh button. If several transformations are applied to a target, the function calls are nested, with the most recent one on the outside. E.g.

lastFunction(secondFunction(firstFunction(original_target)))

To undo a function you just applied, click the target and use the Undo Function. The outermost function will be removed from the target text. You don’t have to use the Apply Function menu - you may also select your target, click Edit, then manually edit the target text by hand.

A quick summary of the Apply Function menu is as follows. Some functions are used on multiple targets (either by selecting multiple entries in the dialog, using wildcards, or both). More information about functions can be found on the Functions page on the Graphite Docs Page.

• Combine

  Combining the data points from multiple metrics using functions such as sum()

• Transform

  Create a new set of data points after applying a transformation such as derivation or timeshifting to the input.

• Calculate

  Perform running calculations on the data points such as a moving average, or forecasts.

• Filter

  Discard, limit or maintain counts/maximums/minimums of the data points.

• Special

  Miscellaneous functions, mostly contain routines that control the appearance of that target on the graph - For example setting a legend name or line color.