[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] DAGMan Errors



Hey Marco,

The version I am using is:

$CondorVersion: 9.0.11 Mar 12 2022 BuildID: 578027 PackageID: 9.0.11-1 $
$CondorPlatform: x86_64_CentOS7 $

To avoid using inline submit, would it be possible to give numerical arguments to my jobs in the .dag file?

For example:

example.dag:

JOB EX1 EX1.subÂN=20 x=1.0 mg=0.05 D=10 l_0=0.5 lambda=100.0 acc=1e-08 ms=100 mem=1

andÂ

EX1.sub:

environment = "JULIA_DEPOT_PATH='/lustre/fs23/group/nic/tangelides/.julia:$JULIA_DEPOT_PATH'"
executable = /lustre/fs23/group/nic/tangelides/julia-1.7.2/bin/julia
arguments = /afs/ifh.de/user/a/angeltak/Schwinger-Wilson-MPS/run_Schwinger_Wilson_dag.jl $(N) $(x) $(mg) $(D) $(l_0) $(lambda) $(acc) $(ms)
transfer_input_files = /afs/ifh.de/user/a/angeltak/Schwinger-Wilson-MPS/run_Schwinger_Wilson_dag.jl
request_memory = $(mem)G
max_retries = 5
should_transfer_files = IF_NEEDED
output = /afs/ifh.de/user/a/angeltak/logs/N_$(N)_x_$(x)_D_$(D)_mg_$(mg)_l0_$(l_0)_$(Cluster)_$(Process).out
output = /afs/ifh.de/user/a/angeltak/logs/N_$(N)_x_$(x)_D_$(D)_mg_$(mg)_l0_$(l_0)_$(Cluster)_$(Process).error
output = /afs/ifh.de/user/a/angeltak/logs/N_$(N)_x_$(x)_D_$(D)_mg_$(mg)_l0_$(l_0)_$(Cluster)_$(Process).log
queue

with

run_Schwinger_Wilson_dag.jl:

using LinearAlgebra
using Arpack
using BenchmarkTools
using Plots
using LaTeXStrings
using Test
using HDF5
include("MPO.jl")
include("variational_first_excited_state_MPS_algorithm.jl")

# ----------------------------------------------------------------------------------------------------------------------------------

# Generate data for entanglement entropy vs mass plot

N = parse(Int, ARGS[1])
x = parse(Float64, ARGS[2])
mg = parse(Float64, ARGS[3])
D = parse(Int64, ARGS[4])
l_0 = parse(Float64, ARGS[5])
lambda = parse(Float64, ARGS[6])
accuracy = parse(Float64, ARGS[7])
max_sweep_number = parse(Int64, ARGS[8])

generate_entropy_data(mg, x, N, D, accuracy, lambda, l_0, max_sweep_number)

# ----------------------------------------------------------------------------------------------------------------------------------

Kind regards,

Takis

On Mon, Apr 25, 2022 at 2:01 PM Marco van Zwetselaar <zwets@xxxxxxxxxx> wrote:
Hi Takis,

On 25/04/2022 14:07, Takis Angelides wrote:
> I did not get a syntax error. I tried what you suggested and got an
> error message: _CONDOR_DAGMAN_USE_CONDOR_SUBMIT=False: Command not
> found. from the command line. after
> submitting:Â_CONDOR_DAGMAN_USE_CONDOR_SUBMIT=False condor_submit_dag
> -f Schwinger_Wilson_MPS.dag in the command line.

I'm afraid I can't help you there, that command looks perfectly fine (at
least on Linux, I have no clue about Windows). However, see below.

> I also tried adding to .dag the line CONFIG dagman.config and in this
> dagman.config file I have the line DAGMAN_USE_CONDOR_SUBMIT = False.
> This also failed to switch this configuration to False.

What version of HTCondor are you using? The DAGMAN_USE_CONDOR_SUBMIT
knob was at some point renamed DAGMAN_USE_DIRECT_SUBMIT (with inverse
meaning, and defaulting to True).

Given that the knob defaults to True, I suppose you shouldn't need to
set anything (as that would be equivalent to setting the other one to
False). However ...

> Do you also have an idea about the second error I am seeing about
> preauthentication failure?

Yes, this looks like its caused by the same bug discussed in this
thread:
https://lists.cs.wisc.edu/archive/htcondor-users/2022-April/msg00031.shtml,
which occurs precisely when using direct submit (which became default in
9.7.0).

What's happening is that HTCondor incorrectly tries to obtain the system
credentials, while it is running as a user (it's doing a "direct
submission") and should be getting the user's credentials.

You could try working around the issue with
DAGMAN_USE_DIRECT_SUBMIT=False, but I suppose that will not work with
your inline job definitions.

Cheers
Marco

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/