Grafana 3.0 Stable Released | Hacker News Reader

Grafana 3.0 Stable Released

grafana.org

207 pointstorkelo34 comments

Here are some release highlights :)

  * Big UI improvements and polish
  * Redesigned plugin architecture and SDK  
  * Grafana-cli command line tool shipped with grafana, 
    installs plugins 
  * Persisted dashboard playlists
  * Preferences like home dashboard, timezone and theme on 
    org and user level
  * New plugin types, Apps, and Panels.
  * New platform site built around Grafana at Grafana.net
    * Plugin repository
    * Dashboard repository (coming soon)
    * Hosted Grafana and Hosted metrics (coming soon)
    * Monitoring and backups of on-prem Grafana (coming soon)
  * Annotation popovers can contain clickable links
  * Templated data source Easiy reuse the same dashboard for multiple data source instances
  * OSX Homebrew support Homebrew installation instructions
  * Support for InfluxDB 0.11+ (and new functions)

Youtube screencast with feature showcase: https://www.youtube.com/watch?v=1kJyQKgk_oY

You (and the other contributors) are a machine. Congratulations on the release!

Thank you!

Things I wish were included in that release:

- Some amount of caching/pre-fetching so moving the graph around doesn't take a few seconds each time. At least in the cases where I am zooming in and all the data is already loaded on the client it shouldn't need a network round trip.

- The ability to drag the charts left and right by keeping the shift key pressed, as is the default on Dygraphs charts (dygraphs.com)

- Switch from Flot to Dygraphs would make client-side rendering faster (at least for the typical line charts)

Grafana is mainly built around time series databases that have good roll ups and are usually quite fast. Graphite queries usually only take 50-150ms (except some rare cases). This is changing a bit with newer slower time series databases that don't have good roll ups or are generally slow.

The main problem with prefetching or doing delta fetches is that some queries depend on the whole time range to execute (like moving average, integral, derivate).

I'm a very happy Grafana user. The move to a server based app worried me at first, but it's been humming along nicely the entire time. It's a shame that graphite is still so hard to install/configure (at least from rhel 6 and lower - things might be better on other distros/versions) though. If you're looking to start making some money as a business, I'd be spending a bit of time making sure the data sources are as easy to stand up as possible.

Congratulations on the new release and, really, fantastic work!

FWIW I've switched from Graphite to InfluxDB as a drop-in replacement (ie, influxdb natively accepts graphite-formatted data, I'm ignoring all the extra nice influx features), and I've found that they're both really easy to set up and keep running together :)

I've been using hostedgraphite.com for some time. YMWV.

No fullscreen mode via a URL still it seems? It's a big missing feature for when you want to setup overhead monitors and the like.

Looks like a nice release! One thing I really miss in grafana and seems like it's not included is alerts.

It would be so damn convenient to have data visualization and alerts on the same system because usually they are strongly related from the user point of view. And, well, one thing less to setup and maintain.

However, I am aware of the debate whether alerts do actually belong to grafana, or should it be responsible for visualization only and seems like they have settled with the later. Which definitely makes sense because once you start to expand to alerting it's a whole new world and I respect the choice. So yeah, I am a bit sad as a user, however I totally get the authors.

Maybe it will be available as a plugin?

That being said... What tools HNers are using for placing alerts on data stored in graphite?

I'd love to hear how others manage it too. I have a bunch of little python scripts in cron jobs that pull / compare numbers from graphite and then post to slack. Adhoc but at least there was nothing much to set up / maintain and it's totally flexible.

As things expand though I'd definitely like to move to something to look after it for me.

You could try Seyren https://github.com/scobal/seyren

We're extremely happy campers here combining the Grafana dashboard with https://prometheus.io/ Datastore and Alerts. That project has some serious traction and is one of just a few that seems actually built for the cloud and distributed systems first, as it's primarily role and not host based.

Can I ask how you manage logs?

My minimal understanding is that prometheus is time-series only so you'd have to supplement with something like ElasticSearch to aggregate logs. Does this mean you are alerting only on metrics or have multiple alert systems or ...?

We run Grafana for production dashboards with a KairosDB (we are a C* shop) and use ELK for text logs. Grafana can add annotations from ElasticSearch, but beyond that we are looking at our ES alerting options.

Interesting. I briefly looked at it previously but have never had the chance to play around.

Did you (or has anyone else) migrate your legacy data from graphite to prometheus? What's the grafana support like on top? Do you need to run a proxy or something to support the same querying or does it require reworking all the existing graphs?

Prometheus developer here.

> Did you (or has anyone else) migrate your legacy data from graphite to prometheus?

Prometheus isn't intended as a long-term data store, so there's usually not much point. The data model is also quite different.

> What's the grafana support like on top?

Grafana supports Prometheus as a 1st class integration, and when I was speaking with the Raintank team in person they are very supportive of Prometheus.

> Do you need to run a proxy or something to support the same querying or does it require reworking all the existing graphs?

The data model and query language is more powerful and very different, you have to redo all the graphs.

To aid transition we have https://github.com/prometheus/graphite_exporter which will take in graphite formatted exports from your clients, and convert it to the format Prometheus likes.

The one thing that holds me back is the pull nature of Prometheus. We are pushing metrics already, so moving to a pull model feels like a return to the 20th century.

It actually tends to work a lot better. You almost never actually need metric granularity so fine that push is necessary. If you do, your usually using a purpose built tool for that. It also saves you from accidentally DOSing yourself when someone starts unexpectedly emitting more metrics than you can ingest (due to bug, more traffic, etc). All doable with push based systems but they tend to end up using queues or something to compensate which is its own kind of pain.

Actually, you probably want to have both. Push is great when you are collecting results of some operations, otherwise you would have to save data somewhere until it is fetched.

The https://github.com/prometheus/pushgateway serves those cases, primarily service-level batch jobs.

We use Sensu with the graphite plugins[1] for alerting based on graphite queries. It works quite well, but you do need to set up Sensu server – which is very straightforward in my experience

[1]: https://github.com/sensu-plugins/sensu-plugins-graphite

I've only had a cursory glance over it but there's Moira for creating alerts based off graphite metrics

http://moira.readthedocs.io/en/latest/

Alerting is in the pipeline according to https://github.com/grafana/grafana/issues/2209

Alerting is on the way :D https://github.com/grafana/grafana/blob/e69658128cd972d94eb8...

I use icinga2 (nagios-like monitoring) to store perfdata/metrics in graphite, then use Grafana to visualize those stats.

I've used grafana from a browser at a company once, and liked it a lot. However, when I've looked at installing it myself, the docs immediately start talking about other projects, like influxdb and elasticsearch, things I know little about nor want to. Conversely, the things I do want to know about, what is it written in, architecture, how to send data?, etc, are missing or buried somewhere.

Is there an easy tutorial on how to start sending it data after install? That's all I want. Also, would like to use standard tools like curl, python, and postgres, is that possible?

Technically speaking, you don't send data to Grafana -- it polls a data source (like influxdb, elasticsearch and graphite).

Ok, that's good information I've never realized, perhaps part of the problem. Or the docs' problem.

It would have been nice to warn beforehand that support for InfluxDB-0.8.X may be broken. I'm having the same issue as https://github.com/grafana/influxdb-08-datasource/issues/3 so if you are still using 0.8.x be careful with this upgrade!

I have some random historical meteorological data (multiple time series, ~3 million values, a couple of parameters, nothing big) lying around in a MySQL database which I would love to explore. Grafana looks like a nice tool to at least visualize it.

Which supported data source would you recommend for a quick test ride?

Not openTSDB. That's by far the most difficult to set up. It's also the best in terms of scalability and performance.

Actually, druid.io pretty handily beats it. I've witnessed (another team member) scale opentsdb to an obscene number of writes, but getting data out with more than 3-4 tags for a near-realtime dashboard was pretty much impossible.

Here is a comment on druid I posted recently:

https://news.ycombinator.com/item?id=11654715

I would go with ElasticSearch.

Reply on news.ycombinator.com