Parallel chains hanging on Ubuntu 18.04

Hi all,

I have happily used Stan without issue for a few years. It’s been a great tool and the hard work that the development team has put into it really shows.

Previously, I have primarily used Stan via pystan from MacOS; however I recently switched to Ubuntu and have found that I am unable to run parallel chains.

When I call my_model.sampling(chains=some_number_gt_one), only a single chain actually runs (I checked pystan’s output in the terminal and the processes running on the computer with htop). Once this single chain completes, the program hangs forever. It is presumably waiting for the remaining chains to complete, which doesn’t happen (they haven’t even started!).

I presume using Ubuntu on servers and the like is a common use-case, so I am wondering if anyone else has had the same problem, and if I need to configure my system to enable the parallel chains.

Some details on my environment in case it is relevant:

$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description:    Ubuntu 18.04.2 LTS
Release:        18.04
Codename:       bionic

And for this specific project:

$ pip freeze
backcall==0.1.0
beautifulsoup4==4.9.0
bs4==0.0.1
certifi==2020.4.5.1
chardet==3.0.4
Cython==0.29.17
decorator==4.4.2
idna==2.9
ipython==7.13.0
ipython-genutils==0.2.0
jedi==0.17.0
joblib==0.11
lxml==4.5.0
numpy==1.18.3
pandas==1.0.3
parso==0.7.0
pexpect==4.8.0
pickleshare==0.7.5
pkg-resources==0.0.0
progressbar2==3.51.0
prompt-toolkit==3.0.5
ptyprocess==0.6.0
Pygments==2.6.1
pystan==2.19.1.1
python-dateutil==2.8.1
python-utils==2.4.0
pytz==2019.3
requests==2.23.0
six==1.14.0
soupsieve==2.0
traitlets==4.3.3
urllib3==1.25.9
wcwidth==0.1.9

That’s unusual. The same program works fine on macOS?

Does it complete with 1 chain?

Yep - it works fine on MacOS and a single chain completes just fine.

Hmmm so I am on
|Distributor ID: Ubuntu
|Description: Ubuntu 19.10
|Release: 19.10
|Codename: eoan

Running with R and rstan and everything works fine. Do you have your model code and some sample data I can play with? I ran a few models in pystan under 15 and 16 but moved over to R from 17 forward.