sophistifunk API to get all nodes is collection and is paginated. Is it happening due to pagination not handled correctly? 

Could be... Seems odd that it would be a problem on a graph with 20 nodes tho?

ekassis can you provide reproducible Jenkinsfike to reproduce this issue?

vivek I added the pipeline code in the description.

I'm going to look into this, see if there's an easy fix we can get in without changing a bunch of stuff that might have to be re-done shortly.

I couldn't get the sample pipeline to run, kept getting weird syntax issues that make me think Jira has mangled it with some hidden characters or something. I cooked up a pipeline with > 100 nodes, and can't seem to trigger any bad behaviour: 

Pipeline: https://gist.github.com/sophistifunk/000b5a14f31f51abb47be2830368c2b6

SS: 

sophistifunk, you can check the attached pipeline code on which we are facing the issue.

Noting that in the example you provided there the no sequential stages inside the parallel nodes.

In addition your code that launch parallel stages is totally different then the one we are using.
Can you please tell us how using your way to have sequential stages inside parallel nodes?

(meta: apparently I can comment here again. Weird.)

I don't think this is related to paging, as this was causing issues in the past (JENKINS-41205) and the solution was to set the limit to 10,000 nodes.

Not sure what else could be causing it, without a pipeline I can run that demonstrates the issue

sophistifunk the pipeline is already attached, please use it

Updated pipeline script triggers the bad behaviour for me also, thanks. I'm looking into the cause.

Progress of a sort, it seems to be a backend issue, not a front-end issue like I first thought. Switching focus there

Righto, good progress, I've built a minimal pipeline that generates the issue reliably, and figured out the trigger: https://gist.github.com/sophistifunk/39cec79eb7d353b67bd70b75a61f95bc

It seems that any time you have all of these conditions:

1. A parallel stage
2. The last parallel branch (in file order) is nested in a sequential wrapper, ie:

   stage('foo') { stages { /*...*/ } }

1. The sequential wrapper only has one child stage

Then the generated graph data is broken for nodes coming after the offending top-level stage.

The problem is (almost certainly) in Blue Ocean server, or (far less likely) the Pipeline plugin. I'll find out which one and get a fix in, should be a small change once I identify the right spot.

PR (waiting on some dogfood issues for building) https://github.com/jenkinsci/blueocean-plugin/pull/1836

Fixed in master @ e49553f1c8792e4e6f74285d3f8377a8f4ec3d99

thank you sophistifunk!

Can you please help me with JENKINS-53735 since noone is replying ?