Dedalus2 Install error

59 views
Skip to first unread message

Peng Deng

unread,
Mar 11, 2026, 4:08:38 AMMar 11
to Dedalus Users
Dear Dedalus users,

Due to the renewal of the HPC facility in my university, I need to re-install Dedalus2 in the new HPC.

I follow the custom conda installation and use the bash file conda_install_dedalus2.sh. 
The sub-environment dedalus2 could be installed, but when I use 'python3 -m dedalus test' to verify the installation, errors pop up. Those errors (see the attached screenshot) seem to relate to multi-thread issue.

Should some options in the bash file (conda_install_dedalus2.sh) be modified accordingly to resolve this issue?

Thanks for your help.

Best,
Peng 
install_error.PNG

Calum Skene

unread,
Mar 11, 2026, 5:35:05 AMMar 11
to Dedalus Users
Hi Peng,
Can you try running the tests again with less workers. Something like
PYTEST_WORKERS=3 python3 -m dedalus test
You are currently using 256 workers which is maybe the cause of the issue (especially if on a login node).
Best,
Calum

Peng Deng

unread,
Mar 11, 2026, 8:18:44 AMMar 11
to Dedalus Users
Hi Calum,

Thank you very much for your reply.

I tried 'PYTEST_WORKERS=3 python3 -m dedalus test', but the tests are still run with 256 workers, same errors pop up, see the attached screenshot.

Best,
Peng

install_error_2.PNG

Calum Skene

unread,
Mar 11, 2026, 12:32:46 PMMar 11
to Dedalus Users
Hi Peng,
Ah okay, PYTEST_WORKERS must be a d3 thing. Can you try running the tests in a job with 4 cores or something small. Hopefully this runs better.
Best,
Calum

Peng Deng

unread,
Mar 18, 2026, 3:21:35 AMMar 18
to Dedalus Users
Hi Calum,

I installed dedalus3 on the HPC facility in my university and found all the tests smoothly passed with the PYTEST_WORKERS.

Now, I am using dedalus3 to simulate quasi-geostrophic vorticity equation in a doubly periodic domain and I am a little confused with the tau method to enforce gauge condition in doubly periodic domain.

The equation I want to simulate is a nondimensional three-layer quasi-geostrophic potential vorticity equation, which reads as:

dt(q_1) + 2*dx(q_1) + [2*alpha/(alpha + 1)]*dx(psi_1) + HD(q_1) = -J(psi_1,q_1)
dt(q_2) + [2/(alpha+1)]*dx(q_2) + [(2-2alpha)/(alpha+1)]*dx(psi2) + HD(q_2) = -J(psi_2)
dt(q_3) + [-2/(alpha+1)]*(1 - slope)*dx(psi_3) + HD(q_3) + drag = -J(psi_3,q_3)
q1 = L(psi1) + (psi2 - psi1)
q2 = L(psi2) + (psi1 + psi3 - 2*psi2)
q3 = L(psi3) + (psi2 - psi3)

where HD denotes hyperdiffusion and drag is parameterized as a quadratic friction.

When I did the simulation in dedalus2, I use the condition 'nx ==0 & ny ==0' to avoid evolving the 'nx==0 & ny ==0' mode but keep this mode zero for both PV and streamfunction.

Now I want to add the tau terms in the equation and prescribe the 'integ(psi_1) = 0', 'integ(psi_2) = 0' and 'integ(psi_3) = 0'.  

Do you have any suggestions on how to add the tau terms in the QG PV equations above?

I attach the code for simulating the above QG PV equation in d2 and d3 respectively, although the tau terms in  d3 code is not added in the equation.

Best,
Peng   
GF20_three_layer_vary_pv_interior_slope.py
GF20_three_layer_vary_pv_interior_slope_d3.py
Reply all
Reply to author
Forward
0 new messages