Grafana 7.0 follows community traces, lays groundwork for next steps

Grafana 7.0 follows community traces, lays groundwork for next steps

When DevClass spoke to Grafana’s VP of product and Prometheus maintainer Tom Wilkie earlier this month, he made sure to point out that 7.0 would be “the first release of Grafana that has metrics, logs and traces in one place”, offering users “to switch from logs to traces really easily”. 

And the open observability platform has done just that with its seventh major release, out today, which sees it finally throwing traces into the mix as well as introducing transformations and math operations for all data sources.

The addition of Zipkin, Loki and Jaeger support comes at the right time, with tracing slowly catching on in cloud native circles, and visualisation requests along those lines slowly trickling in. However, the capability was initially “driven by our own desires,” Wilkie told us, since the Grafana team, pretty much like the one building Prometheus, sees itself as one of its biggest users.

“And we’re really focusing, at least initially, on a kind of instant response workflow. We really want to make these kinds of transitions quick and easy. So when you get paged [at] 4am in the morning, you can do all of this in a bleary eyed state, and really kind of reduce that mean time to recovery.”

According to Wilkie, the new setup is meant to prevent people from unnecessary “copying and pasting and kind of manual translating”, which was needed in earlier versions and sometimes led to spending too long on the wrong thing.

Another highly requested feature that finally made its way into the 7.0 release is a “rich query history” that also allows users to go through old queries and annotate them. The UI team, which is behind the enhancement, has been generally busy for this release, since it also comes with a redesigned UI for editing panels and a unified data model. The latter supposedly gets rid of disparate configurations that led to some grief in earlier versions by providing consistent options and behaviours throughout all panels.

Speaking of which, Grafana now also offers a panel inspector, “which allows you to dive into any panel in any query and see what happened – you can see the raw data, you can see timing information, but on a data source by data source basis” to get more information of what happened when. “We’ve done this for Loki as well”, Wilkie added, “you can go and see, what indexes were hit, and how much parallelisation was applied to the query”.

Enterprise users can also look forward to more insight on dashboard usage to find out which ones maybe aren’t needed anymore or look into a dashboard’s history to learn who made which modifications. Other new enterprise features allow users to see who else is looking at a dashboard, which can for instance help with coordination when working on incidents in that 4 o’clock scenario, and Okta/SAML team syncs.

While many still tend to think of Grafana only in combination with Prometheus, this isn’t exactly what Wilkie has in mind. Of course, there are integrations with Jaeger, Elastic, Graphite, and Zipkin as well, but this isn’t the end of the line. “In the future, we want to expand this so that everyone can have this experience. For our users, it’s about owning their own observability strategy – it’s about deciding what the right combination of technologies is for them and Grafana enabling that decision”.

To make that happen, a lot has changed under the hood, with Grafana for example now sporting an Apache Arrow-based data and plugin framework. It’s also the basis for new additions, such as the data transformation pipeline ops teams can use to join, filter, and pivot data, which has the added bonus of allowing simple maths operations across queries.

Another step into the “anti-consolidation” direction, as Wilkie calls it, is the addition of CloudWatch logs as a data source for Grafana, which was developed in partnership with Amazon. What comes next, is really up to the user base, if you ask Wilkie. Currently, the IoT space, such as monitoring wind turbine activity, and BI seem to peak customer interest here and there. And while this isn’t something the next release will definitely hit on, the team will definitely investigate it further, laying the groundwork for “the next big thing”.