Ok thats enough for me


#1

You’d better bear with me…

Ive been using Grafana for very long time (2+ years) and its been ok (not good, just ok).
With every release its becoming worse and worse; one bug causes cascade of other bugs, compatibility issues, alerting issues etc etc.

Just to give you an example from recently (today in fact):: In dashboard for Azure, I needed Network in/out visualization, I added heatmap, nicely configured it to my liking, have not touched alerting(in heatmap-type there is no such tab), preview is nicely generating, click Save and guess what??

Error pops up saying invalid alert data. Just total facepalm.

Am testing paid solution for monitoring (not Grafana) and will soon (in the matter of hours) switch to it completely. $15/year is not much. But definitely worth having service that just works

Sorry Grafana - its time to say goodbye. Its been nice to be here


#2

What paid solution are you going to?


#3

Ive tested two of them in fact :: site24x7.com -> nice, clear UI, but they are simply lying saying that price starts from $9,99 but they force you to buy additional staff (it gets added to every plan you select) and in reality your minimum monthly bill is $20.

And second one I tested was DataDog.com. Plans starts from $15/host/year (for cooperate agreement) and its real price (= without any hidden costs). For $15 you get dashboards like Grafana, + real alerting + complex log management (filtering, piping etc) + lot more useful features…

I bought DataDog…


#4

As the creator of Grafana & Project lead this makes me sad. We work really hard to make fundamental improvements while still providing a full backward compatibility and stable releases. We have not had any other reports that Grafana has gotten more buggy over time, if that was the case we would have taken stock and tried to address that. The error you have received with the heatmap panel and Azure is very strange, no other reports like it & unable to replicate it here which is indeed very strange. Will ask our Azure & Heatmap plugin devs to investigate more.

Are there any other instances where you feel the quality of Grafana has gone down? You mention compatibility issues, alerting issues etc etc. Any details there would be helpful.


#5

First :: many thanks for replying to me.
Second :: What did I mean by saying Greafana gets more buggy from release to release? Lets talk facts:: In the latest Grafana there is a bug when you add new panel to dashboard it gets added but when you click on panel title and want to edit it, whole panel randomly disappears -> this leaves “shadow” of this panel -> this causes error on both saving dashboard and on loading dashboard (empty space with red exclamation mark) -> this is undeleteable and unreplaceable (=you cannot place other panel over the “shadowed” one).
This is perfect example of how one bug can start cvascade of other bugs.

Another (cascading) bug :: when you hover mouse over graph (doesnt matter its type) to display tooltip -> its all ok, but data update while tooltip is displayed causes tooltip text to disappear, leavinjg tooltip frame on the screen… When you move your mouse out of tooltip area, the tooltip (or better named:: its background) stays on screen -> this stops whole graph from refreshing.
Whats interesting is that this can be observed on Azure dashboard only. Other dashboards are ok as of tooltips.

As of alerting :: its only available with graphs. Why? Whyt cannot we set alerts for, say, heatmap or other types?

Incompleteness of UI:: Add graph type table, go to Edit it > open Column Styles pane. There is a group called Type? Type of what?
As we are with graph type table… another cascading bug:: Cloumn Styles pane. Click +Add link on the left and set Type to String, configure value mappings and than switch Type to whatever else, configure and switch Type back to String… you will notice that preview stays on what you’ve changed Type to…

Another bug is within importing of dashboards :: when you copy&paste JSON in text-area and click Import it sometimes generates error saying:: Unable to import. Importing very same code, but within the file - using Upload button works perfectly. Strange

Also - explain to me whats the rationale behind restricting alerting only to graph instead of allowing alerting from any type of visualisation? Also what the hell is alerting icon on the left panel for, if - when clicked - loads page that will not let you add alerts?
Speaking of it:: when you create an alert, save it, click on Alert icon on the left pane, your alert is visible there, but its colored orange with exclamation mark; once clicked you are sent to corresponding graph with red error msg saying that (hold yourself) :: The datasource does not support alerting queries
How come graph let me set alert, when it does not support it? And why there was alert tab/pane visible when its unsupported??
Not to mention there is no way of correctly saving alert on “unsupported” graph - it will leave “shadowed” alert and this will render as red exclamation marks instead of graph…

By ‘cascading’ I mean bug(s) that triggers either other bugs or unwanted behavior.
All of the above bugs leads to problematic/being unable to/corrupted saving dashboard…

OK, thats everything Ive noticed. Whats sad is that Grafana once was great software. It is no more. There are others who produce better monitoring software.
Many of the above bugs have been here from 2-3 releases. It feels like you (=developers in general; not you personally) stopped taking care of quality…

To be fair :: what I really liuke about Grafana is very simple, well-organised UI…well done here.

EDIT:: Just (moment ago) finished moving every production-grade monitors to DataDog. Will leave my current account (with free Hosted Grafana and just standard CPU/Network monitors) just to see how Grafana is doing)…

EDIT2:: Just noticed:: when you select specific region on one visualisation (regardless its type), what gets enlarged (zoomed in)? Every visualisation on open dashboard.


#6

This is a completely new an (up unitl now) unreported bug. Unable to replicate it in latest stable & master.

Can you replicate it here? https://play.grafana.org/dashboard/new?orgId=1 What version of Grafana & Browser & OS are you using to replicate this bug?

This is also a new unreported bug that we have not heard before. Maybe because Azure is yet not that common a plugin. Can you open a bug report on github.com/grafana/grafana with instruction on how to replicate it?

Because we have not yet added this feature. There have been a million other things to add & improve.

Type of style to apply.

Ok, not sure I 100% understand the issue but does not sound like a critical bug (not that it shouldn’t be fixed).

Since we did the first iteration of Alerting the core Grafana team has been focusing on other core Grafana features (UX & dashboard grid & missing visualizations and reviewing & refactoring thousands of community pull requests & bug reports). There has been no big PRs & community enhancents to alerting in the mean time so the feature has not improved much since it’s inception.

This error message should be there from the start when you create the alert (it is for me). Only a limited core set of data sources support alerting since they are executed by the backend and old datasource plugins where frontend only (js). We have since then added support for backend plugins so now external plugins can also add alerting support but this requires more work for each data source.

I am not sure what “unsupported” graph means. We have no similar reports to what you describe so not sure what your experiencing.

Thanks for reporting this. We have no other reports like it so this is very valuable. Can you open a github issue with server & chrome network logs for when you get this corrupt dashboard that wont save / load?

We care a great deal and are very thankful for any bug reports that lets us know that things are not working as they should.

Not sure what you mean by this sentance. When you zoom in a region on a graph you change the time range on the dashboard so all other panels will show data for the new time range. This is how Grafana (and 99% of other dashboards) work. Is this unexpected to you or do you mean something else?


Integration with Prometheus Alertmanager
#7

As of my contribution (by bug reports or otherwise):: I will not do it, simply because of not having time to…
Ive migrated out of Grafana, because it stopped being suitable for what I (we as a company) do…

This sentence summarized it all - if for you alerting is not important than its just hight time to start ignoring Grafana. Firstly because things like alerts are absolute core of every monitoring engines, and secondly because you are starting doing something and leave it halfway finished due to other things that (in your opinion) needs to be done. And with other things will be the same. Generally - with this kind of attitude from someone as important for project as you are - spending time working on Grafana is just waste of time. It will never be done properly.

Read above…
Either implement something fully or not implement at all.

Grafana OK
Other dashboards NO
In DataDog when I have many graphs omn dashboard and I make a selection with mouse, what get enlarged is the selected area only (new window with enlarged selection opens leaving dashboard unchanged).

Edit3:: what the hell is that? Why there are placeholders instead of actual texts/data?

Hey what do you base this assumption on? In fact there were plenty of community enhacements, but they were blocked by team members. Why? Answer is simple and well known to you!
See screenshot below: Why there are so many issues regarding alerting closed?
Capture6

You said that there wasnt any community enhancement? Thats simply a lie. There were many, but - for some reason - you (+team) decided to just ignore them.

EDIT:: Please look at this. Will only say that it was you who authored linked issue…back on 6th of December 2016, so 1,5 year ago. And whats the current state of implementation of this?

Next one:: what the hell is Playlist and is this really needed/core funtionality to show off? Does any serious/mature monitoring solution have this feature? NO they dont

Next one::

Requirements
There are no dependencies

and

Dependencies
Go 1.8.1
NodeJS LTS

So, no dependencies, or two of them?

Whats your input on all of above?


#8

I’m glad that you have found your new monitoring tool, which fits better your needs. But I don’t understand why this angry childish theater here. That’s really not a necessary.

I guess you don’t understand:

  • Grafana is a visualization tool - it’s not a real monitoring, but it’s trying to be for some use cases (you have still a plenty options for monitoring https://github.com/monitoringsucks/tool-repos)
  • your Grafana needs are probably different than needs of majority people (for example I like Grafana graph zoom functionality, but you don’t)
  • you are just consumer and your are not contributing - obvious reason “not having time to” - that’s not how you can make this open source better

Sorry mate, but your score is :-1: (because your language) and Grafana team :+1:
Please be more civil next time (you already have received this request from he community).


#9

Since when voicing concern(s) is childish?

As per your points:: I clearly see (and understand) what Grafana is trying (with fluctuating success-rate) to be. Visualisation should also be backed up by alerting…imagine situation where, say, Mr X (employer of IT dep, whose taskm is to monitor production env. of huge multinational company)…installs Grafana, configures dashboards (all the heatmaps, etc) and just log out. And now - when he is logged out of Grafana, doing something else, BAM, production env stops, nobody knows why, nobody knows … anything. Chaos.

Now lets change scenario a little bit - once marker hits warning level, Grafana sends an email alert, Mr X notices it, acts accordingly, saves nerves and prevents disaster

So you clearly see from the story above that alerting is crucial in any kind of visualizing software, doesnt matter if its real-monitoring-focused or not.

I dont see how I was/am not civil. May you show me examples?

EDIT::

Seriously?!? Dont even try to teach me how to contribute/make OSS better… I have done this many times (probably more than you and @torkel alltogether), but Im very choosy which project(s) I contribute to…
I usually dont contribute to projects of low-quality or ones that have arrogant team…


#10

@jangaraj fact that - since your last reply here - you have been online and havn’t replied to me speaks for itself…
Not to mention it confirms what Ive suspected…


#11

So you have only one host to monitor? Also, your Data has to live on Datadog hosted servers correct?


#12

No, we have 100+ hosts (physical) + circa 100VMs…
As of data living on DD servers:: no. Its up to you where you store your data that feeds monitors…


#13

I’m not sure if this is similar to what @dev11 is saying here, but I know there is poor error handling when it comes to panel plugins not functioning properly. This is one instance that has been irritating with “unsupported” plugins (meaning that they weren’t developed by the Grafana team). Basically, if there is an error within the panel, the panel won’t render including the menu option. Therefore, you can’t even remove it. Space is just taken with an empty, resizable hole.