Analysis starts slowly with CP3.0.0



I was trying CP 3.0.0 noticed it takes quite a long time before the actual analysis gets started. It doesn’t feel particularly slow in test mode, or once the processing gets going. So I’m wondering about the delay:

Starting workers on address tcp://
Progress Counter({u'Unprocessed': 36})

# after 6-7 minutes...

Worker 1: Starting job
Progress Counter({u'Unprocessed': 35, u'InProcess': 1})
Worker 1: Doing job: 1
Worker 1: Wed Mar 28 17:25:54 2018: Image # 1, module Images # 1: CPU_time = 0.0
0 secs, Wall_time = 0.00 secs
Worker 1: Wed Mar 28 17:25:54 2018: Image # 1, module Metadata # 2: CPU_time = 0
.00 secs, Wall_time = 0.00 secs

Is there something I could do to debug what’s going on?



Are you using the LoadData module? I’ve had similar issues using large imagesets and LoadData.


Hmm, yeah, 6-7 minutes is a long time to wait!

You only have 36 image sets, so it’s not that; are the files especially big or have an especially large number of channels? Do you get this delay every time, only with certain pipelines, or only sporadically?


Pipeline doesn’t seem to matter, it’s the same with a pipeline with only IdentifyPrimaryObjects. I’m using the new default input modules.

This seems to be about CP process start up in general. It takes about 35s for the GUI to show up. When I tested with maximum number of processors set to two, progress started at about 1min 15s; with 4 processors, at about 2min 30s. And the ~7min corresponds with having all 12 threads in play.

Does that make sense? Should it be faster? The PC has Intel® Xeon® CPU X5670 @ 2.93GHz and 24 GB of RAM. CP 2.2.0 doesn’t have this delay.


I definitely think it should be faster, so I’ve filed a bug report; if our software engineers have questions for you or want you to try something to confirm, that link will be where.

So sorry for the inconvenience!