Creating Grafana Dashboards for Prometheus | Grafana Setup & Simple Dashboard (Chart, Gauge, Table)

hi I'm Julius so maybe you just got started with Prometheus monitoring and now you're wondering how to create dashboards based on the data you're collecting so that's where most people turn to the open source grafana dashboard Builder since Prometheus itself doesn't actually come with full dashboarding features that's why in this video I'm going to give you a quick intro to using grafana with Prometheus first we're going to download and run grafana just in case you don't have it yet then we're going to configure it to fetch data from a Prometheus server and finally we're going to create a demo dashboard there's a billion things we could do with grafana but just to get started we're going to create three types of panels a Time series chart a gauge panel and a table by the way you don't need a Prometheus server to follow along for this video I'm also just going to be using the publicly available promenaps demo server that you can reach at demo.promlabs.com if you do want to use your own server and you don't have one yet check out this setup video I made up here alright let's go [Music] first let's set up grafana there's many different ways how you can do this but two of the basic alternatives are to either run it using Docker or by using the pre-built binaries that you can run directly on your machine if you want to go the docker route you can start grafana using their public Docker image like this starts grafana and makes it listen on Port 3000 you can also run it in the background using the Dash D flag give the container a proper name and a persistent data volume so that it remembers the dashboards that you build across restarts the second alternative is to use the pre-compiled binaries that graphina offers for that let's head to the graphina homepage at grafina.com and go to the downloads page and instead of cloud we're going to select self-managed and finally we're going to click on the download button on grafana itself I'm going to download the Open Source version although it doesn't really matter as long as you're only using the free features it already pre-selected Linux as the operating system here which is great since I'm on Linux now I'll just scroll down to the Standalone binary section down here and click on the archive name to download it now I'll switch to a terminal here and go to the downloads folder and I'll unpack the tarball and change into the unpacked directory let's have a look at what got unpacked here so the conf directory has the default configuration files for grafana which started on Port 3000. and the bin directory has the grafana server binary which we can use right now to start grafana using the default config all right grafana is now listening on Port 3000 so let's head there in the browser the default username and password are admin and admin so we can just log in with that now it asks me to change the admin password but for this demo I will just skip that as a first step before we can even start visualizing any data from Prometheus we need to First tell grafana about our Prometheus server and the URL that it's running on to do that I'm going to click on add your first data source here and select Prometheus as the data source type we only need to change two Fields here first I'm going to choose demo Prometheus as the data source name and since this is the first data source that we are creating it's already set to be the default and secondly as the URL of this Prometheus server I'm going to use the prom Labs demo Prometheus server that is running publicly on the internet there are many other settings here like HTTP and authentication options but for this demo we can leave all of them at their default values since that primitive server is not secured in any way so now we scroll down and click save and test to check that the data source is working great so now grafana is all set up to send prom KL queries to Prometheus so let's create our first dashboard by going to the dashboards menu on the left and selecting new dashboard that drops us into an empty new dashboard where we can add panels like time series charts tables histogram heat maps and many other visualization types so let's start out with just a Time series chart showing some data from our Prometheus server I'll click on add a new panel which puts us into the edit view for that new panel now there's a lot going on here but there's roughly three main sections at the top there's the actual visualization itself that will end up showing on the dashboard below that is where you input the queries to fetch the data to display above and on the right you can find a large number of options to configure what the panel should look like all right so let's first set the time range for the graph to 30 minutes since that'll look better with the data that we're going to be selecting now let's add a query below we can see that our demo Prometheus data source is already selected here since it was the default data source so we can leave that as is now we can use this graphical promql query Editor to select the latency histogram metric from a demo service that the Primitive server is monitoring and rather than showing that metric in its raw form let's add some transformation steps to it to aggregate it and calculate the 95th percentile latency from it for every path and Method label and don't worry if you don't get all the prompt L details here we'll definitely talk about those in later videos by the way in case you're finding this video useful and want to see more Prometheus content in the future please consider liking the video and maybe even subscribing to the channel because that really helps out the YouTube algorithm and more people will be able to find these videos so now let's run the query and you can see that the panel above shows the fetch time series data as a chart over the last 30 minutes by the way grafana has two different modes for inputting prompt ql queries you can either use this graphical query Builder or you can switch over to the code view to see and edit the textual prompt URL query itself note that this generated query now contains a grafana template variable that is not part of promptql itself those template variables are replaced with concrete values by grafana before sending a query to Prometheus so in this example this rate interval variable helps us automatically select the rate averaging window that is suitable for the resolution of the promql query if you look at the legend in the graph and hover over the series you'll see that grafana has automatically named the series as Prometheus wood with label pairs in curly braces if we want to format the legend differently we can open the query options down here and change the legend option from Auto to custom that allows us to enter a custom Legend format string where we can interpolate label values by referring to the label names in double curly braces so for example let's change the legend to only show the label values for the path in method labels and separate those by long Dash now our Legend already looks nicer when looking at a panel on the dashboard it's important to be able to quickly understand what the panel shows and what it means so let's give this a panel a title of 95th percentile request latency and we can see that the panel now shows this title another thing that's not clear yet by just looking at the panel is the unit we might know that the underlying histogram metric is in seconds but grafana has no clue about that yet so let's search for unit in the panel options on the right and select seconds to tell grafana that our underlining data is in seconds now grafana interprets our data correctly and automatically shows values in milliseconds on the y-axis finally let's hit the apply button which takes us back to the dashboard with our newly created graph panel okay let's create a gauge panel next by hitting the add panel button and going to add a new panel again and this time we're going to set the panel type to gauge and we're going to enter a promptql query that shows us CPU usage for three different demo service instances with the output values varying somewhere between 0 and 1 for each of the instances we can explicitly tell grafana about these Min and Max values so that it knows how large to make the big value bar in each gauge display and let's also set the unit to be a percentage but a percentage that is actually coming in as a ratio between 0 and 1. now that looks a whole lot better already with a percentage being displayed in the panel now we can also add color coding to the different value thresholds so let's say that we want to consider 80 CPU usage high and we want to display everything above that as red for that let's search for the thresholds options up here and change the default red threshold of 80 to 0.8 since that's the equivalent of 80 in our underlying ratio data all right now we see where the red region would begin and finally let's clear the search bar up here and give the chart a title again we're going to call it instant CPU usage and hit apply to save this panel as our third and last panel let's display some data in a table so we're going to create the new panel as before but this time we're going to set the panel type to table in our prompt ql query we're going to select the top three request rates that our demo service is getting grouped by path and Method labels Now by default this table won't look right yet we see a bunch of different values over time for a single selected Time series and we can select which time series to show but we would actually like to see all time series at once and only the latest value for each one of them so first of all let's switch the format option from time series to table and secondly let's change the query type from range to instant so that we only fetch One latest value for each series if we run the query again now the table looks much better already showing us the top three request rates however maybe we don't want to show the timestamp column since that timestamp is always going to be the latest one so there's actually multiple ways to get rid of that and I'll just show you two alternatives the first option is to add a query transform by going to the transform section and choosing the organize Fields transform in there we can simply click the I icon next to the time field to hide it from the table now I'll just undo that and also remove the transform again so I can show you the other option for this other option we have to scroll down all the way in the panel options add a field overwrite and then choose field width type for that override now that allows us to select a specific type of field to overwrite and we're going to select time we then add an override property to hide this kind of field in the table and finally I still need to enable the overwrite we just created now as you can see it removed the time field in our table again and finally let's again choose a unit for our panel here which in this case is requests per second or RPS and we'll set the panel title to be top three rates great now we have a dashboard with three panels as a quick last thing I'll show you how you can organize panels within a dashboard into thematic rows for example we might want to group together the two HTTP stats panels into one row and the CPU stats into another one we are going to do that by clicking add panel as before but instead of adding a new panel we're going to add a new row instead let's repeat that one more time and now we can rename the rows by hovering over them and hitting the settings icon the first one we're going to call HTTP stats and the second one we're going to call CPU stats now we can grab each panel by its title bar and drag it into the right row you can collapse the rows by clicking on them and you can also drag them around using the right hand side handles to reorder them on the dashboard finally let's save the dashboard by hitting the save icon and I'm going to call it demo dashboard now we got the dashboard showing a chart a gauge and a table and we organized our panels into rows alright so that's how easy it is to get started with grafana and Prometheus I hope this was useful and I'm going to cover a lot more Prometheus topics in future videos that go beyond these Basics so be sure to like And subscribe if you want to get more of that also if you would like to learn Prometheus in a structured way right now I'm making these self-paced training courses at training.promlabs.com so check out those as well and see you soon

Channel: Prometheus Monitoring with Julius | PromLabs

Views: 34,277

Keywords: prometheus, monitoring, grafana, dashboarding, time series chart, gauge panel, table panel, graph, data source, tutorial, howto, guide

Id: EGgtJUjky8w

Length: 13min 51sec (831 seconds)

Published: Tue Jan 03 2023