the lack of custom colorization is a big one for me as well. It makes the graphs and visual data much easier to understand at a quick glance than trying to easily differentiate between the greens and blues.
Still on KB3. (Though I have a test instance of KB4 up.)
Interface on 4 is more complicated. Wasn't clear how to move over a large amount of my dashboards. Ran into some instances of features missing in 4.
Still on Kibana3
Kibana4 is still lacking a few features that is essential when you are using it to quickly dig through lots of information to filter out noise.
Not sure. I have a large set of saved dashboards. So one day I decided: lets try to recreate all of these in Kibana 4. I started running into problems doing so. I found a few bugs for the features I needed... and then quickly gave up.
It's a lot of work that I'm not looking forward to do, and finding out that all the features weren't present anyways quickly made me give up on it.
I have not spent that much time so far in Kibana4 against the same data we use Kibana3 for. One of the first pain points I found was the need to go from Discover to Visualization to Dashboard for every little thing you want to add to a Dashboard. It is better now that you can skip the Discover part. Another is you can't seem to assign colors to a query like in Kibana3. For example I have a dashboard for IIS logs, I have 4x queries that search give me data for each host producing the data, they are all blue tinted. Then I have another 4x queries that get response times, 0-999,1000-1999,2000-3999 and 4000-1000000, they are tinted green,yellowgreenish,yellow and red. Then I have a few histograms that use either only the host queries or only the response time queries. The host queries are mostly informational but the response time ones are there to grab your attention.
I have mostly used Kibana4 with some earthquake data, totally different use case but I definitely can see that Kibana4 shines when exploring data.
I created an issue here, https://github.com/elastic/kibana/issues/4060 for one of the problems I found regarding tables.
One feature that i would like to be able to do, but isn't necessarily missing since I never found a way to do this in kibana 3, is the ability to "overlap" a graph. essentially i would like to overlap a graph of application response times with a graph of apache requests so a direct correlation could be made on a single graph.
Basically such as this:
Nobody on our team has been able to migrate to Kibana 4 yet due to a variety of issues. For me personally, it's editable filters.
Follow-up question: for those of you who still use Kibana 3, have you made any improvements or fixed any bugs?
I'm wondering whether it would make sense to start a community supported fork of Kibana 3 where we could all collaborate on bug fixes and upgrades that would make use of newer Elasticsearch features.
K3 has the ability to put "other" in a piechart, to account for everything other than say the top 10. Lack of this feature in K4 is preventing complete adoption. Please add this.
Unfortunately, this is a capability that was present in Elasticsearch Facets, analytics framework used by K3, that is missing in Elasticsearch Aggregations, their new analytics framework used by K4. So, Kibana can't add it until this exists in ES: https://github.com/elastic/kibana/issues/1961
I am with @elvarb. Been using Kibana3 (and ES) for 2-3 years. Kibana 4 is a complete change. We analyse in excess of 120m docs per day, realtime on Mysql, and the issues surrounding the change in operation of K3 -> K4 are huge.. for our use case.
You cant edit a filter
You cant set aliases, and colors for queries (easily)
You cant set timespans through a Histogram.. then delete the filter when done.
You cant seem to have multiple queries making up one visualisation... easily.
The time it takes to build a datebased histogram is pitiful, compared to K3. I have 6 dashboards on K3, running multiple queries... blindingly fast. One on K4 and that dies, and K4 is on a different machine.
You cant...... and so on and so on.
The geomapping in K4 is streets ahead of K3 (thanks guys), but suspect this s down to Leaflet.js .. so we will be deploying both.
Essentially it looks like K4 is trying to mimic splunk, and is built for a static dashboard, not for a scenario where you use graphs to get to the 'pain points' of systems.. getting queries to make sense.. so you can fix the pain points.
what it does seem to do is provide a 'static set of data' to show what you are doing, but this is not the use most people would use.. I suspect.
Going back through Discover to add an filter is just too difficult and time consuming.
Please keep Kibana 3... as it si will be fine.. just dont break the link. We will use K4 for where it is good, but I dont see why the major change.... it makes no sense to me.
I think you can do that by sing one index, but with multiple graph types. We will be overlapping apache with our live mysql logs soon (as soon as I get a production cluster up), and intend to do that by putting Apache logs as a seperate _type in the index. So queries on _type. Only possible (currently) I believe with K3.
I was wondering if anyone had forked Kibana 3 myself. If I remember some of the beta posts that led up to the Kibana 4 release, doesn't Kibana 4 have a different focus than Kibana 3? Um, something like the two versions differ in how they approach analyzing data.
If that's true, wouldn't it make sense to keep Kibana 3 around? Since people are currently happy with that approach?
As for my personal situation, I know I found Kibana 3 a LOT more intuitive. But I never built much on it before K4 was announced.
That said, I do really want https://github.com/elastic/kibana/issues/1583 editable filters. I'd also really really like to be able to export and import visualizations and dashboards. As in some code I could stick in a gist along side notes on how to make sure the data in ES matches the Kibana configuration. That would let the community share their work.
Another thought, has anyone at Elastic thought about making an importer for those who have hundreds of K3 dashboards? Something that can at least pull in the basics and save some time?
we are using Kibana3 mainly for analyzing log data. Unfortunately there are still lots of features missing in kibana4 that prevent us from migrating:
- missing vector maps (country-code based)
- colors cannot be adjusted
- no possibility to export/import dashboards (to/from files) to exchange them between different elasticsearch clusters / indexes
- no possibility to easily create an identical copy of a dashboard that uses data from a different index
- no editing of filters
- no manual adding of (querystring) filters
- no folding/hiding of panels on a dashboards
- no naming ("legend value") of queries
- no "missing field" and "other values" in Top Terms Aggregation (Facets in K3)
- in any aggregation, order by name doesn't seem to be possible
- no possibility to change the resolution of a date histogram temporarily in the dashboard directly
- no possibility to change visualization style (terms: bar/pie/table; histogram: bars/lines/points) without creating a new visualization
@tbragin We spoke today at the meet up. For the "other" pie-slice issue, I was asking if you could use the "sum_other_doc_count" field in the response json. Here's a small snip:
@Sesha_Narahari Thanks for the pointer, I'll get back to you!
The most cause is the incompatibility between K3 and K4, I cannot recreate all my dashboard.
I understand the 'developer' motivations, but from the point of view as product it was the worst path.
The other are
- lack of black theme
- lack to choose color for series
- lack of panel 'map'
There was talk of forking Kibana3. Did anyone set one up? We've made a number of fixes to it internally, and have no problem sharing the fixes.
We agreed with all of the above reasons listed for not upgrading to Kibana4.
Once the major issues are fixed, we will upgrade, but until then, it would be good to maintain K3.
@tinle I would love to see your fixes pulled into the fork of Kibana 3!
Here's the github repo that we are tracking: https://github.com/kibana-community/kibana3
I pulled in the latest code from the official branch and some community submitted improvements like csv exports.
We can share ownership of that repo with you to make it easier for you to commit your changes.