[puneet336]

Hi,
i am using version 15.0 of fluent on centos6 with intel mpi

time -p fluent -g 3d -t 16 -i jour.jou -mpi=intel
i get :

sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument

also i tried with:
time -p fluent -g 3d -t 16 -i jour.jou -mpi=openmpi
and again i get same message.

i ran HPL on the node and performance was normal. It was then that i noticed that wheneaver sched_setaffinity() call failed, my job experiences degraded performance.Ans also it writes out some files.
(journal.jou)
/solve/set/time-step 0.0005
/solve/dual-time-iterate 100000 150
and even after 24 hours, i cannot see even a single cas/dat file (maybe due to the slowness??).

Is there any command line argument/switch via which i can provide "valid argument" to sched_setaffinity via fluent ?

please let me know if more information is needed from my side.

[LuckyTran]

Try running on fewer threads, e.g. -t 8
If that works...

Check your hosts file to make sure you have 16 entries.
Check your mpi

You can manually set affinity using set-affinity

To automatically set affinity use:
(set-affinity "")

To manually set affinity use:
(set-affinity "node-id1 processor1 node-id2 processor2 ...")

e.g. if your machine host file is
machine-1 machine-1 machine-2 machine-2
(set-affinity "0 0 1 1 2 0 3 1")

will specify
node0 to run on processor-0 of machine-1
node1 to run on processor-1 of machine-1

node2 to run on processor-0 of machine-2
node3 to run on processor-1 of machine-2

[sina_mech]

Originally Posted by puneet336

Hi,
i am using version 15.0 of fluent on centos6 with intel mpi

time -p fluent -g 3d -t 16 -i jour.jou -mpi=intel
i get :

sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument
sched_setaffinity() call failed: Invalid argument

also i tried with:
time -p fluent -g 3d -t 16 -i jour.jou -mpi=openmpi
and again i get same message.

i ran HPL on the node and performance was normal. It was then that i noticed that wheneaver sched_setaffinity() call failed, my job experiences degraded performance.Ans also it writes out some files.
(journal.jou)
/solve/set/time-step 0.0005
/solve/dual-time-iterate 100000 150
and even after 24 hours, i cannot see even a single cas/dat file (maybe due to the slowness??).

Is there any command line argument/switch via which i can provide "valid argument" to sched_setaffinity via fluent ?

please let me know if more information is needed from my side.

Could you solve this problem? I am facing the same issue. I'd appreciate if you share your experience.

[LuckyTran]

Originally Posted by sina_mech

Could you solve this problem? I am facing the same issue. I'd appreciate if you share your experience.

It's trying to set the process affinity for the thread to the job scheduler. Meaning it wants the process to have priority over other tasks in cpu and memory. But it fails, which means you don't have priority and that's why your calculation runs slow. I'm not familiar with CentOS so I can't help you further. The problem isn't Fluent, it's something in your environment.


I've never seen this command time fluent. Is that CentOS specific?

[sina_mech]

It is a CentOS 7. If it is not Fluent, then there should be something that our HPC admins can do to fix it. I will keep searching.