cellprofiler / docker Goto Github PK

• Git
• OpenJDK
• Python
• wget

• bioformats
• Cython
• email
• hd5py
• matplotlib
• nose
• numpy
• pdb
• Pillow
• pyreadline
• pysql
• pyzmq
• readline
• scipy
• tifffile
• wxPython

Travis CI runs a Makefile that runs CellProfiler on the ExampleHuman pipeline and compares this to a gold standard output. However, Travis is failing for 2 reasons.

1. The example has changed at cellprofiler.org.
2. CellProfiler 3.0.0 cannot recreate the output of the old pipeline. There is not an equivalency between the output of IdentifyPrimaryObjects between CellProfiler 2.x and CellProfiler 3.0.

I think the easiest way to solve this problem is to update the gold standard.

Hi all,

We routinely run ~10,000 headless instances of Cellprofiler in parallel on AWS. Thank you guys for creating a docker image! It's the one we've been using.

In about 1 / 1000 runs the docker container would not exit cleanly. We determined that it was likely the docker zombie reaping problem. In short, there were subprocesses in the container that were dead but weren't being cleaned up.

Fortunately, phusion solved this problem and created phusion/baseimage. Their image has a lighweight my_init script which detects and kill zombies. We rebuilt the CellProfiler docker image using phusion/baseimage as the starting point, and haven't seen any zombies with the new image.

The fix is just a 2 line change. If you think this would be useful, I can create a PR.

-- JD

Hello,

At Recursion we've developed an updated, fully-tested version of CellProfiler's Docker from 2.2.0 that also solves the zombie reaping problem. I've tested it in both batch and GUI mode and it works great. I also have a branch open for alpine development since the base is Ubuntu 16.04 and therefore likely much larger than necessary.

Are you interested in a pull request?