{"slug": "grafanacon-2026", "title": "GrafanaCon 2026", "summary": "The article is a sarcastic, first-person account of the author's experience at GrafanaCon 2026 in Barcelona. Despite claiming to \"hate\" the conference, the author details that it was incredibly well-organized, featured excellent speakers and a great venue, and inspired him to explore new Grafana features. The piece also provides background on the author's company, OpenNMS, which was an early adopter and third-party data source for Grafana.", "body_md": "I recently attended my first GrafanaCon, held in Barcelona, Spain, from 20-22 April.\nI hated it.\nTL;DR; There was a lot to hate about this conference. It was incredibly well organized, which made me feel bad about the past events in which I’ve been involved. On the second day I didn’t get my choice of meal, and they never had enough diet soda on hand. The venue was excellent, and the single track nature of the presentations meant I was exposed to amazing speakers with no other options. The worst part is that it made me want to give up my weekends to play with the new version of Grafana. The nerve.\nGrafana was started in 2014, and the earliest references I can find to our involvement with it at OpenNMS are from the latter half of 2015. I know we were really excited about it as a visualization tool, especially since we were working on a new time series data storage library called Newts and we needed a way to display the collected values.\nOpenNMS was also one of the first third-party data sources created for Grafana, and Jesse White spoke about it at the first GrafanaCon. We had limited budget so Jesse or someone on his team ended up representing us at most of the Grafana events, and it took until now for me to attend one.\nIt was, truly, a very good conference. When I was verifying the venue location I noticed that it was also sold out, with over 650 people in attendance.\nThe conference spanned three days. Monday consisted of workshops, with the main conference running Tuesday through Wednesday. The agenda was packed, with keynotes starting at 09:00 and running until 18:00 both days. That said, there were plenty of breaks to enable the hallway track, and the Tuesday evening event (hosted by AWS) allowed even more time for folks to interact.\nThe venue was Palau de Congressos de Catalunya, which is several miles from the Barcelona city center, next to the Universitat Politècnica de Catalunya. Barcelona is my favorite city on the planet, but it can get crowded with tourists. This location was convenient and near a lot of nice restaurants while avoiding the crowds.\nThe most of the activity took place on the main floor, which featured a large area containing the exhibit hall as well as the main auditorium. The stage itself was huge, but the AV setup meant that everyone had a good view. I don’t remember any major issue with the technical side of the sessions and the conference WiFi was actually pretty good as well. The lower level had a large room for breakout sessions and was where lunch was served.\nI did not attend the optional workshops on Monday, but I did use the time to get my badge and meet up with some folks.\nIn my role at AWS I work with a lot of the partner managers for customers, and my main contact at Grafana is Will Nunes. He was working the badge desk and it was nice to finally meet him face to face, and he would end up taking care of me for the time I was there.\nThe main event started on Tuesday. I arrived around 08:30 to find a very nice spread of food and drinks (coffee, juice, water) laid out for folks, and this would be a theme throughout the two days. You really won’t go hungry at a GrafanaCon.\nSince I knew I was going to write this (long) post, I got a seat early, up near the stage. The auditorium had the best seats in my memory - super comfortable with large pull out tables, perfect for laptops and with access to power built-in to the seat backs.\nThe opening keynote featured Raj Dutt and Torkel Ödegaard. To use an Apple analogy, Raj is Steve Jobs to Torkel’s Woz. They showed a cool little video featuring Grot (the GRafana robOT) in Barcelona. I usually have a visceral, negative reaction to GenAI videos, but they used actual footage of Barcelona and just used AI to superimpose Grot on the screen. It was really cool when they showed pictures of Park Güell and Grot became mosiac-tile Grot.\nAfter the video they showed a slide that really incapsulated the Grafana message: openness. It has been the key to their success.\nThey also showed a cool progression of the movement of Grafana across the Gartner “magic quadrant” for observability platforms. Over the past three years there has been a steady progression as Grafana has started to challenge established players, ending with 2025 where they had the strongest “completeness of vision” of the sample.\nThe next speaker was David Kaltschmidt, VP of Engineering. He introduced Grafana 13. We would get a deep-dive into the newest version on Wednesday, but he covered the three main areas of improvement: make it easier to get started, make it scale, and make it available everywhere.\nPoyzan Taneli then spoke on Loki, an open source logging platform maintained by Grafana. The main improvement here was an integration with Apache Kafka. Kafka is a messaging bus with very high performance. We used it with OpenNMS and what was cool is that it was very easy to publish data and then just subscribe to those streams of interest. The new changes will allow Loki to scale more and perform better in those situations that require it.\nThe next speaker was Ted Young, who co-founded OpenTelemetry.\nOpenTelemetry is a CNCF project that provides a framework for telemetry data, such as traces, metrics and logs. It is designed to provide a common API for the instrumentation of applications and infrastructure. It came along about the same time I was leaving OpenNMS and stepping away from observability, but I have been casually following the project since it was announced. I love the fact that this provides a common way for observability platforms to consume telemetry data without relying on a proprietary format or vendor.\nYoung’s presentation was focused mainly on stability, and while it is not exactly sexy, the goal is to make OTel stable by default.\nNo modern conference would be complete without discussing generative AI. Mat Ryer kicked off that discussion (and he would end up serving as the Master of Ceremonies for most of the conference).\nThere were two main areas to cover with respect to AI: the new AI assistant, and using Grafana for AI observability.\nSven Großmann talked about Grafana Assistant. One thing to understand about Grafana is that their are three main versions: Open Source, Enterprise and Cloud. Both the Open Source and Enterprise versions are designed to be run “on prem”, and Grafana Cloud is a hosted option. Cloud has a rolling release cycle, and a lot of features hit it first, such as the new AI powered Assistant.\nGrafana gets a lot of AI stuff right, and having this Assistant makes it easier for users to get the most out of the application. Data science is hard, and that can make data visualization difficult as well. For example, if you are polling for a value periodically, what happens when that poll fails? Usually applications store a null value (NaN) but then what do you do when you try to take an average? Do you treat the unknown values as truly null or do you treat them as zero? For example, if I have a 1GB/sec circuit that goes down, and for an hour I only get one sample, if I treat the missing values as unknown the average bandwidth for that hour would be reported as 1GB/sec, even though no traffic moved through it for most of that time.\nOf course you can tell Grafana to treat those unknowns as zeroes, but it takes at least one extra step. With Grafana Assistant you could just tell the agent how you wanted those values to be evaluated, and it would do the work for you.\nThis will be very useful to both experienced and new Grafana users, and the big announcement was that it was now available in Open Source and Enterprise.\nThe feature does require a Cloud account, but they have a free tier that should provide enough credits for most people to use it. I did ask about why there wasn’t an on-prem version and the answer was simply time to market. They wanted to get this feature out as soon as possible, and it just wasn’t feasible to build it to run locally.\nI get that. At OpenNMS is was always nice when a support customer gave us ssh access into their system, as it was quicker to troubleshoot issues when we could go “hands on”. I can’t imagine the challenges in building a local version of this assistant, from finding a model that would perform well with a license suited for local use, to determining the minimum hardware requirements and then troubleshooting the issues that are sure to arise.\nThe second big “AI” thing was observability applied to AI platforms and applications. Grafana will allow you to instrument you AI applications and agents to capture every LLM call, token usage and latency. This ability to follow conversations, agents and generations in real time will help provide a better user experience while keeping an eye on cost.\nThe final presentation of the morning took a slight detour. Back in May of 2025 there was a change made to a GitHub Action on the Grafana repo that allowed for an outside party to access certain environment variables, including credentials. Instead of pull-request\nan action was uploaded that changed it to pull-request-target\n, which was not safe.\nThe presentation was given by Nick Moore and David Andersson, and it was cool to see how this was caught and fixed. I also learned about “canary” credentials. Since most attacks like this are automated, when a bad actor gets access to credentials they are likely to explore just what kind of access they provide. By creating “dummy” credentials that should never be used, it is trivial to set up an alert when and if they are. It is an easy way to get a warning that something is wrong.\nThis was followed by the first morning break, and I would spend the next couple of hours talking to people around the expo hall.\nMy first stop was the AWS booth which featured “Adventure Quest”. Using a laptop, attendees were given the chance to solve an old skool text adventure game with an observability focus, and winner got a sword (not a real sword but an inflatable one so they could take it home through the airport).\nIt was being run by Guillermo Ruiz who works for AWS out of the Madrid office. It is always fun to meet cool coworkers at conferences, as we probably would never meet otherwise.\nIn the center of the Expo Hall there was a big round “Ask the Experts” area, and I introduced myself to several of the “Grafanistas” working there. Grafana is a fully remote company (well, due to legal reasons they have a physical office in Japan) but there was a large contingent of folks from Munich. I also got to spend some time talking with Ted Young, who was dealing with a large electrical spike that hit his house back home. As someone who travels a lot I can attest that things are most likely to go wrong at home when you are away.\nI then went downstairs for lunch. Like many conferences the meal was served in a box, but unlike other conferences the food was really good for “in a box” type fare. The main course was a “poké bowl” with a fish, meat and veggie option, but then you got an additional sandwich, plus fruit, plus chips … it was pretty tasty. There was so much food I didn’t envy the presenters that afternoon since everyone would be sleepy and the chairs in the auditorium were really comfortable.\nI was late to the afternoon sessions and came in during the end of a presentation by two engineers from LEGO, Paul Farver and Lorenzo Setale.\nI always enjoy it when you get to hear about tech choices made by a company that is also a household name, and for once I was okay with people using LEGO figures in their slides (grin)\nThe next presentation was from Marko Bachvarovski and Bejal Lewis on Alloy. Alloy is a telemetry collector that is meant to replace separate agents with one configurable pipeline. As one might imagine, OpenTelemetry plays an oversized role and this talk covered the roadmap for OTel integration with Alloy.\nI missed the next talks because I was backstage getting ready for my few minutes of fame. I had been asked to represent AWS, and I was to go on just after th", "url": "https://wpnews.pro/news/grafanacon-2026", "canonical_source": "https://www.adventuresinoss.com/grafanacon-2026/", "published_at": "2026-04-29 18:22:00+00:00", "updated_at": "2026-05-24 06:06:26.432618+00:00", "lang": "en", "topics": ["open-source", "developer-tools", "data", "enterprise-software"], "entities": ["GrafanaCon", "Grafana", "OpenNMS", "Newts", "Jesse White", "Barcelona"], "alternates": {"html": "https://wpnews.pro/news/grafanacon-2026", "markdown": "https://wpnews.pro/news/grafanacon-2026.md", "text": "https://wpnews.pro/news/grafanacon-2026.txt", "jsonld": "https://wpnews.pro/news/grafanacon-2026.jsonld"}}