Understanding the debug graph

Hey guys, I’d like to understand the debug graph better and just have a few questions for clarification.

This is a PMV comfort map that I’m currently running:

  1. It has been running for 21min now and it is still not finished which is making me think something went wrong. My grid has 2041 points and the sensor count is the default 200. I was therefore expecting to see 11 branches in the graph but there appears to be only 1?
  2. How can it be that downstream icons have a green tick while their parent is still running?
  3. Assuming the graph is read from top to bottom, I’m wondering why the ‘run-energy-simulation’ branch and the ‘run-irradiance-simulation’ branch are disconnected. I thought one should be feeding into the other?

Update: The CheckJobStatus now says “Completed” but the LoadAssets component is “Unable to download file”:

The web app also says it has “Completed” but the run status says “Failed”:

The debug graph is a bit of a spaghetti monster. I’ve clicked on several of the “X” components in the lower layer but none of them have logs:

Job ID: 4f5ac5eb-1c1d-5511-8dab-4426d4f2684a

Hey Max,

We are in the process of updating how we manage per-account compute resources now that we are in Public Beta which has caused some issues with run stability and with the results. Sorry for the poor experience!

I’m going to ping @mostapha and @antoinedao who might be able to provide more insight on your question 1.

For number 2: the graph should really be read as going forwards in time from top to bottom. That is, the older nodes are higher, the younger ones are lower. So, in your screenshot create-rad-folder has a green check because all it had to do was create a directory. However, its parent run-irradiance-simulation is still waiting on create-octree (among others) which could not start until after create-rad-folder succeeded. It’s more of a “depends on” relationship than a “parent/child” relationship.

For number 3: the debug graph can only render nodes that have actually been scheduled on the cluster and which are reporting a status. So, while the run is executing, the graph can be incomplete as all of the nodes may not yet be executing or providing a status. I can’t quite tell from your screenshot of the completed run, but it looks like those two subgraphs did eventually connect.