Hi, I'm fairly new to Kibana so I apologise if I get terminology wrong or mixed up in this question.
I want to create a Histrogram-esque chart, in which each bar/block in the chart represents a job that has been submitted to a HPC cluster. I want the width of each bar/block to be the difference in time between the start time and the end time of the job (two date fields every job record has), and the height of each block to be the number of GPUs used.
Essentially, the area of each block will represent the number of GPU Hours used by job. I do expect some overlap in blocks (i.e. a job may start during another job, so will have a start time in between the start time and end time of the previous job).
I've uploaded an image illustrating what I want. So essentially, I want to use 4 fields, "JobID", "StartTime", "EndTime", and "NoOfGPUs". Is this possible or is it too complicated?
This exact chart may be difficult with Kibana visualizations- the x axis bars are typically grouped by a single field. I'd recommend crunching gpu hours prior to indexing and then plotting that on the y axis if possible.
Thanks for the reply Jon. I've now taken that approach. But it would've nice to be able to do something like above as it would be a good way to visualise the usage of the HPC machine. Is this something you could consider feeding back?
reading over this, this is actually pretty advanced, and cannot be done with Kibana-Visualize now.
The biggest issue is the dynamic width of the bars, that may or may not overlap. All our other Kibana-visualizations have fixed-width bucket-size, so it would be difficult to shoehorn this in the existing Kibana-code.
Depending on how familiar you are with programming JS and the Elastic Stack in general, you could consider writing a plugin for this:
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.