Grafana Time-Sequence Dashboards with the Rockset-Grafana Plugin

What Is Grafana?

Grafana is an open-source software program platform for time sequence analytics and monitoring. You possibly can join Grafana to a lot of knowledge sources, from PostgreSQL to Prometheus. As soon as your knowledge supply is related, you need to use a built-in question management or editor to fetch knowledge, and construct dashboards out of your knowledge supply. Grafana is incessantly deployed for all kinds of use circumstances, together with DevOps and AdTech.

At Rockset, we primarily use Grafana for monitoring our manufacturing methods, in addition to for DevOps functions. We observe all kinds of metrics, from the variety of question errors to the CPU utilization of our manufacturing machines. Every time a graph deviates from a predefined band of anticipated values, we set off an alert which might hook up with one thing like a PagerDuty integration that will ping an on-call engineer.

Why Construct a Plugin?

As energy customers of Grafana ourselves, we had floated the concept of constructing a Rockset connector for Grafana for a very long time. Due to the realtime nature of Rockset as an operational analytics engine, we believed {that a} Grafana plugin may very well be a superb match for a variety of issues and queries. We realized that we may start monitoring plenty of time sequence metrics that will enable for better transparency into our engineering practices (by monitoring the heartbeat of our GitHub commits into grasp, for instance), in addition to our inside methods that we’re monitoring via Rockset (reminiscent of occasions in our Kubernetes cluster). Another excuse a Rockset-Grafana plugin is useful is as a result of an software developer can use normal SQL to fetch any type of knowledge via Rockset. Lastly, it was one thing that our prospects had beforehand expressed curiosity in. Taking these factors under consideration, constructing a Grafana connector appeared like an apparent and helpful software of Rockset to boost an already highly effective instrument.

How To Construct A Grafana Connector

To construct a working Grafana connector, one must implement a group of Typescript strategies, in addition to a customized consumer interface for retrieving knowledge out of your given datasource. After the plugin has been carried out and take a look at circumstances written, it’s reviewed by the Grafana maintainer staff and built-in into the official checklist of plugins.

The performance that any Grafana connector must implement is:

1. Datasource Specification

   When constructing a plugin, you have to first truly have the ability to fetch the info you’ll be setting up dashboards out of. This typically entails having the consumer specify an API key, password, or database connection URL to fetch the info from.

2. Customized Question Interface

   As soon as a datasource has been specified, a consumer wants to have the ability to question that datasource. Within the case of Rockset, this concerned implementing a customized question editor in HTML and AngularJS that’s proven to the consumer when they’re making a dashboard with Rockset.

3. Question Execution via the API layer

   After the consumer has typed in a question, the info itself wants to truly be fetched and handed to the visualization layer in a really particular format. This entails speaking with the frontend via the consumer’s question modifying, in addition to question execution via the Rockset API and post-processing of outcomes such that they’re handed to the visualization within the correct timeseries format.

Constructing the Rockset-Grafana Plugin

Going again to the steps outlined above, the very first thing that I wanted to do when constructing out the Rockset Connector was to truly join the Rockset Datasource. I constructed out a kind that allowed a consumer to specify the title of the plugin, in addition to the Rockset API key. This concerned constructing out the shape on the frontend, in addition to writing a testDatasource methodology that validated the correct API key with a take a look at question to the Rockset backend via a fast name to the /v1/orgs/self/customers/self/apikeys endpoint within the Rockset API that ensured the API key itself was legitimate.

As soon as the important thing was validated, it was time to construct out the question editor. Within the case of Rockset, now we have to permit a consumer to kind in arbitrary SQL to any of their collections. Moreover, you will need to present informative error messages for syntactically invalid queries or if a consumer is querying on a group that doesn’t exist.

I carried out the question editor with a debounce operate that allowed a consumer to kind their question, then pause so it may very well be executed via the Rockset API. The queries are checked for validity on the backend, and the error is handed to the consumer on the frontend to allow them to obtain an informative error message. Moreover, Grafana requires a timeseries column if you wish to categorical the info by way of an over-time graph. The Timeseries column field permits a consumer to specify a column of their SQL outcomes that they select to pivot their graph axes on. The Format as field is a straightforward dropdown that enables a consumer to specific a Rockset question as a timeseries or as a desk, and this modifications the formatting of the info handed to the graph layer.

After a question has been typed in, validated, and executed, the info is obtained by the Grafana connector. Sadly, we can’t merely go the info to a desk or graph and show it within the Grafana dashboard. We have to extract the user-specified timeseries column, convert it into Unix seconds, and go an array of JSON objects into the visualization layer of Grafana. We are able to additionally neatly counsel the timeseries column if a consumer specifies just one column that’s of kind datetime.

Lastly, as soon as all the question and validations steps have been accomplished, it’s now potential for a consumer of the plugin to visualise their knowledge, and we instantly set about doing that after the plugin had completed being developed.

Use Circumstances and Future Work

As soon as our plugin was full, we began to make use of it for fascinating queries at Rockset. One factor we began out taking a look at was our inside GitHub metrics. Particularly, we began trying on the variety of open points each hour, the variety of closed points and the variety of recordsdata added or modified throughout the course of a day in our firm.

We additionally started monitoring metrics just like the variety of Kubernetes occasions in our dev cluster for higher understanding outages and utilization spikes.

These queries are only a few examples of how Rockset can be utilized with Grafana to offer realtime insights into arbitrary collections of knowledge, and we’re excited to roll this plugin out extra broadly and see how our prospects use it. To see a extra detailed view of the plugin and to get began utilizing it, try the documentation.