Non-homogeneous distribution of processes among nodes using OpenMPI - cluster-computing

I am running my executable with OpenMPI on a cluster using the SLURM resource managing software. I would like to find a way to specify how many and which processes should be assigned to each of the nodes, where the number of processes might be different for each node.
An example to clarify what I am looking for: Suppose I want to run 7 processes on 3 nodes. Then I want to be able to say something like:
node 1 should run the process with rank n, node 2 and 3 should each run 3 of the remaining processes.
I do not care which physical node is node 1, as all the nodes are equal on the cluster I am using. Also I do not know a priori which nodes I will get assigned by SLURM, so I cannot hard-code the nodes' names in a hostfile. An example in the OpenMPI documentation that I found would define the hostfile like this for my example:
aa slots=1
bb slots=3
cc slots=3
but I have two problems with this approach:
I do not know a priori the names aa, bb, cc of the nodes.
Even if I knew them, the process on node aa does not necessarily have the right rank.

Thanks to Hristo Iliev's comment, I found the solution for the example stated in the question:
#!/bin/bash
HOSTFILE=./myHostfile
RANKFILE=./myRankfile
# Write the names of the nodes allocated by SLURM to a file
scontrol show hostname ${SLURM_NODELIST} > $HOSTFILE
# Number of processes
numProcs=7
# Number of nodes
numNodes=${SLURM_JOB_NUM_NODES}
# Counts the number of processes already assigned
count=0
while read p; do
# Write the node names to a rank file
if [ $count == 0 ]
then
echo "rank $count=$p slot=0-7" > $RANKFILE
let count=$count+1
let numNodes=$numNodes-1 # Number of nodes that are still available
else
# Compute the number of processes that should be assigned to this node
# by dividing the number of processes that still need to be assigned by
# the number of nodes that are still available. (This automatically "floor"s the result.)
let numProcsNode=($numProcs-$count)/$numNodes
for i in `seq 1 $numProcsNode`
do
echo "rank $count=$p slot=0-7" >> $RANKFILE
let count=$count+1
done
let numNodes=$numNodes-1 # Number of nodes that are still available
fi
done < $HOSTFILE
mpirun --display-map -np $numProcs -rf $RANKFILE hostname
It is a bit ugly though. And probably "slot=0-7" should not have the "7" hard-coded.

Related

Problems with Orca and OpenMPI for parallel jobs

Hello to the community:
I recently started to use ORCA software for some quantum calculation but I have been having a lot of problems to lunch a parallel calculation in the cluster of my University.
To install Orca I used the static version:
orca_4_2_1_linux_x86-64_openmpi314.tar.xz.
In a shared direction of the cluster (/data/shared/opt/ORCA/).
And putted in my ~/.bash_profile:
export PATH="/data/shared/opt/ORCA/orca_4_2_1_linux_x86-64_openmpi314:$PATH"
export LD_LIBRARY_PATH="/data/shared/opt/ORCA/orca_4_2_1_linux_x86-64_openmpi314:$LD_LIBRARY_PATH"
For the installation of the corresponding OpenMPI version (3.1.4)
tar -xvf openmpi-3.1.4.tar.gz
cd openmpi-3.1.4
./configure --prefix="/data/shared/opt/ORCA/openmpi314/"
make -j 10
make install
When I use the frontend server all is wonderful:
With a .sh like this:
#! /bin/bash
export PATH="/data/shared/opt/ORCA/openmpi314/bin:$PATH"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/data/shared/opt/ORCA/openmpi314/lib"
$(which orca) test.inp > test.out
and an input like this:
# Computation of myjob at b3lyp/6-31+G(d,p)
%pal nprocs 10 end
%maxcore 8192
! RKS B3LYP 6-31+G(d,p)
! TightSCF Grid5 NoFinalGrid
! Opt
! Freq
%cpcm
smd true
SMDsolvent "water"
end
* xyz 0 1
C 0 0 0
O 0 0 1.5
*
The problem appears when I use the nodes:
.inp file:
#! Computation at RKS B3LYP/6-31+G(d,p) for cis1_bh267_m_Cell_152
%pal nprocs 12 end
%maxcore 8192
! RKS B3LYP 6-31+G(d,p)
! TightSCF Grid5 NoFinalGrid
! Opt
! Freq
%cpcm
smd true
SMDsolvent "water"
end
* xyz 0 1
C -4.38728130 0.21799058 0.17853303
C -3.02072869 0.82609890 -0.29733316
F -2.96869122 2.10937041 0.07179384
F -3.01136328 0.87651596 -1.63230798
C -1.82118365 0.05327804 0.23420220
O -2.26240947 -0.92805650 1.01540713
C -0.53557484 0.33394113 -0.05236121
C 0.54692198 -0.46942807 0.50027196
O 0.31128292 -1.43114232 1.22440290
C 1.93990391 -0.12927675 0.16510948
C 2.87355011 -1.15536140 -0.00858832
C 4.18738231 -0.82592189 -0.32880964
C 4.53045856 0.52514329 -0.45102225
N 3.63662927 1.52101319 -0.26705841
C 2.36381718 1.20228695 0.03146190
F -4.51788749 0.24084604 1.49796862
F -4.53935644 -1.04617745 -0.19111502
F -5.43718443 0.87033190 -0.30564680
H -1.46980819 -1.48461498 1.39034280
H -0.26291843 1.15748249 -0.71875720
H 2.57132559 -2.20300864 0.10283592
H 4.93858460 -1.60267627 -0.48060140
H 5.55483009 0.83859415 -0.70271364
H 1.67507560 2.05019549 0.17738396
*
.sh file (Slurm job):
#!/bin/bash
#SBATCH -p deflt #which partition I want
#SBATCH -o cis1_bh267_m_Cell_152_myjob.out #path for the slurm output
#SBATCH -e cis1_bh267_m_Cell_152_myjob.err #path for the slurm error output
#SBATCH -c 12 #number of cpu(logical cores)/task (task is normally an MPI process, default is one and the option to change it is -n)
#SBATCH -t 2-00:00 #how many time I want the resources (this impacts the job priority as well)
#SBATCH --job-name=cis1_bh267_m_Cell_152 #(to recognize your jobs when checking them with "squeue -u USERID")
#SBATCH -N 1 #number of node, usually 1 when no parallelization over nodes
#SBATCH --nice=0 #lowering your priority if >0
#SBATCH --gpus=0 #number of gpu you want
# This block is echoing some SLURM variables
echo "Jobid = $SLURM_JOBID"
echo "Host = $SLURM_JOB_NODELIST"
echo "Jobname = $SLURM_JOB_NAME"
echo "Subcwd = $SLURM_SUBMIT_DIR"
echo "SLURM_CPUS_PER_TASK = $SLURM_CPUS_PER_TASK"
# This block is for the execution of the program
export PATH="/data/shared/opt/ORCA/openmpi314/bin:$PATH"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/data/shared/opt/ORCA/openmpi314/lib"
$(which orca) ${SLURM_JOB_NAME}.inp > ${SLURM_JOB_NAME}.log --use-hwthread-cpus
I used the --use-hwthread-cpus flag as a recommendation but the same problem appears with and without this flag.
All the error is:
There are not enough slots available in the system to satisfy the 12 slots that were requested by the application: /data/shared/opt/ORCA/orca_4_2_1_linux_x86-64_openmpi314/orca_gtoint_mpi
Either request fewer slots for your application, or make more slots available for use. A "slot" is the Open MPI term for an allocatable unit where we can launch a process. The number of slots available are defined by the environment in which Open MPI processes are run:
1. Hostfile, via "slots=N" clauses (N defaults to number of processor cores if not provided)
2. The --host command line parameter, via a ":N" suffix on the hostname (N defaults to 1 if not provided)
3. Resource manager (e.g., SLURM, PBS/Torque, LSF, etc.)
4. If none of a hostfile, the --host command line parameter, or an RM is present, Open MPI defaults to the number of processor cores In all the above cases, if you want Open MPI to default to the number
of hardware threads instead of the number of processor cores, use the --use-hwthread-cpus option.
Alternatively, you can use the --oversubscribe option to ignore the number of available slots when deciding the number of processes to launch.
*[file orca_tools/qcmsg.cpp, line 458]:
.... aborting the run*
When I go to the output of the calculation, it looks like start to run but when launch the parallel jobs fail and give:
ORCA finished by error termination in GTOInt
Calling Command: mpirun -np 12 --use-hwthread-cpus /data/shared/opt/ORCA/orca_4_2_1_linux_x86-64_openmpi314/orca_gtoint_mpi cis1_bh267_m_Cell_448.int.tmp cis1_bh267_m_Cell_448
[file orca_tools/qcmsg.cpp, line 458]:
.... aborting the run
We have two kind of nodes on the cluster:
A punch of them are:
Xeon 6-core E-2136 # 3.30GHz (12 logical cores) and Nvidia GTX 1070Ti
And the other ones:
AMD Epyc 24-core (24 logical cores) and 4x Nvidia RTX 2080Ti
Using the command scontrol show node the details of one node of each group are:
First Group:
NodeName=fang1 Arch=x86_64 CoresPerSocket=6
CPUAlloc=12 CPUTot=12 CPULoad=12.00
AvailableFeatures=(null)
ActiveFeatures=(null)
Gres=gpu:gtx1070ti:1
NodeAddr=fang1 NodeHostName=fang1 Version=19.05.5
OS=Linux 5.7.12-arch1-1 #1 SMP PREEMPT Fri, 31 Jul 2020 17:38:22 +0000
RealMemory=15923 AllocMem=0 FreeMem=171 Sockets=1 Boards=1
State=ALLOCATED ThreadsPerCore=2 TmpDisk=7961 Weight=1 Owner=N/A MCS_label=N/A
Partitions=deflt,debug,long
BootTime=2020-10-27T09:56:18 SlurmdStartTime=2020-10-27T15:33:51
CfgTRES=cpu=12,mem=15923M,billing=12,gres/gpu=1,gres/gpu:gtx1070ti=1
AllocTRES=cpu=12,gres/gpu=1,gres/gpu:gtx1070ti=1
CapWatts=n/a
CurrentWatts=0 AveWatts=0
ExtSensorsJoules=n/s ExtSensorsWatts=0 ExtSensorsTemp=n/s
Second Group
NodeName=fang50 Arch=x86_64 CoresPerSocket=24
CPUAlloc=48 CPUTot=48 CPULoad=48.00
AvailableFeatures=(null)
ActiveFeatures=(null)
Gres=gpu:rtx2080ti:4
NodeAddr=fang50 NodeHostName=fang50 Version=19.05.5
OS=Linux 5.7.12-arch1-1 #1 SMP PREEMPT Fri, 31 Jul 2020 17:38:22 +0000
RealMemory=64245 AllocMem=0 FreeMem=807 Sockets=1 Boards=1
State=ALLOCATED ThreadsPerCore=2 TmpDisk=32122 Weight=1 Owner=N/A MCS_label=N/A
Partitions=deflt,long
BootTime=2020-12-15T10:09:43 SlurmdStartTime=2020-12-15T10:14:17
CfgTRES=cpu=48,mem=64245M,billing=48,gres/gpu=4,gres/gpu:rtx2080ti=4
AllocTRES=cpu=48,gres/gpu=4,gres/gpu:rtx2080ti=4
CapWatts=n/a
CurrentWatts=0 AveWatts=0
ExtSensorsJoules=n/s ExtSensorsWatts=0 ExtSensorsTemp=n/s
I use in the script of Slurm the flag -c, --cpus-per-task = integer; and in the input for Orca the command %pal nprocs integer end. I tested different combinations of this two parameters in order to see if I am using more CPU than the available:
-c, --cpus-per-task = integer
%pal nprocs integer end
None
6
None
3
None
2
1
2
1
12
2
6
3
4
12
12
With different amount of memories: 8000 MBi and 2000 MBi (my total memory is around 15 GBi). And in all the cases the same error appears. I am not an expert user neither in ORCA non in informatic (but maybe you guess this for the extension of the question), so maybe the solution is simple but I really don’t have it, Idon't know what's going on!
A lot of thanks in advance,
Alejandro.
Faced the same issue.
Explicit declaration --prefix ${OMPI_HOME} directly as ORCA parameter and using of static linked ORCA version helps me:
export RSH_COMMAND="/usr/bin/ssh"
export PARAMS="--mca routed direct --oversubscribe -machinefile ${HOSTS_FILE} --prefix ${OMPI_HOME}"
$ORCA_DIR/orca $WORKDIR/$JOBFILE.inp "$PARAMS" > $WORKDIR/$JOBFILE.out
Also, It's better to build OpenMPI 3.1.x with --disable-builtin-atomics flag.
Thank you #Alexey for your answer. And sorry for the wrong Tag, like I said, I am pretty rookie on this stuff.
The problem was not in the Orca or OpenMPI configuration but in the bash script used for scheduled the Slurm job.
I thought that the entire Orca job itself was what Slurm call a "task". For that reason I declared the flag --cpus-per-task equal to the number of parallel jobs that I want to do with Orca. But the problem is that each parallel Orca job (that is launch using OpenMPI) is a task for Slurm. Therefore with my Slurm script I was reserving a node with at least 12 CPU, but when Orca launch their parallel jobs, each one ask for 12 CPU, so: "There are not enough slots available ..." because I needed 144 CPU.
The rest of the cases in the table of my Question fails for another reason. I was launching at the same time 5 different Orca calculation. Now, because --cpus-per-task could be None, 1, 2 or 3; the five calculation might enter in the same node or in another node with this amount of free CPU, but when Orca ask for the parallel jobs, fail again because there are not this amount of CPU on the node.
The solution that I found is pretty simple. On the .sh script for Slurm I putted this:
#SBATCH --mincpus=n*m
#SBATCH --ntasks=n
#SBATCH --cpus-per-task m
Instead of only:
#SBATCH --cpus-per-task m
Where n will be equal to the number of parallel jobs specified on the Orca input (%pal nprocs n end) and m the number of CPU that you want to use for each parallel Orca job.
In my case I used n = 12, m = 1. With the flag --mincpus I ensured to take a node with at least 12 CPU and allocated them. With the --cpus-per-task is pretty evident what this flag do (even for me :-) ), which, by the way, has a default value of 1 and I don't know if more than 1 CPU for each OpenMPI Orca job improve the velocity of the calculation. And --ntasks gives the information to Slurm of how many task you will do.
Of course if you know the number of task and the CPU per task is easy to know how many CPU you need to reserve, but I don't know if this is easy to Slurm too :-). So, to be sure that I allocate the correct number of CPU i used --mincpus flag, but maybe is not needed. The thing is that it works now ^_^.
It is also important to take into account the amount of memory that you declare in the input of Orca in order of do not exceed the available memory. For example, if you have 12 task and a RAM of 15000 MBi, the right amount of memory to declared should be no more than 15000/12 = 1250 MBi
I had a similar problem with parallel jobs before. The slurm also output not enough slots error.
My solution is to change parallel threads into parallel processes. For my system is to change
#SBATCH -c 24
into
#SBATCH -n 24
and everything works just fine.

optimize parallelisation in SLURM cluster: the case of genome alignemnt

I would like to understand what is the best way of using bwa in parallel in a SLURM cluster. Obviously, this will depend on the computational limits that I have as user.
bwa software has an argument "-t" specifying the number of threads. Let's imagine that I use bwa mem -t 3 ref.fa sampleA.fq.gz, this will mean that bwa split the job on three tasks/threads. In other words, it will align three reads at a time in parallel (I guess).
Now, if I want to run this command on several samples and in a SLURM cluster, Shall I specify the number of tasks as for bwa mem, and specify the number of CPUs per task(for instance 2)? Which would be:
sbatch -c 2 -n 3 bwa.sh
where bwa.sh containes:
cat data.info | while read indv; do
bwa mem -t 3 ref.fa sample${indv}.fq.gz
done
Do you have any suggestion? Or can you improve/correct my reasoning?
With -c 2 you are asking to have 2 CPUs per task.
With -n 3 you are asking to have 3 tasks.
That configuration prepares a set of resources that comprises 6 CPUs in up to 3 different nodes. But your script only used 3 CPUs (-t 3), so you are wasting resources and probably using resources that does not belong to you (because the task will use 3 CPUs and you only asked for 2 CPUs per task).
For that specific script, -c 3 is the proper parameter (the other defaults to one task):
sbatch -c 3 bwa.sh

How to share memory in cluster machine (qsub openmpi)

dear all!
I have a question about sharing memory in cluster. I am a new to cluster, and fail to solve my problem after trying about several weeks, so I look for help here, any suggestion would be grateful!
I want to use soapdenovo, a software that was used to assemble human genome to assemble my data. However, it failed in one step because shortage of memory (the memory is 512G in my machine). So I turned to cluster machine (which have three big nodes, each node have 512 memory too), and started to learn submit job with qsub. Considering that one node couldn't solve my problem, I googled and found that openmpi may help, but when I running openmpi with demo data, it seemed it only run the command several times. Then I found to use openmpi, the software must include library of openmpi, and I didn't know whether soapdenovo is support openmpi, I had asked the question but the author didn't give me answer yet. Suppose soapdenovo support the openmpi, how should I solve my problem. If it didn't support openmpi, can I use memory in different nodes to run the software?
The problem had tortured my so much, thanks for any help. Following is what had I do and some information about the cluster machine:
Install openmpi and submit the job
1) The script of job:
#!/bin/bash
#
#$ -cwd
#$ -j y
#$ -S /bin/bash
#
export PATH=/tools/openmpi/bin:$PATH
export LD_LIBRARY_PATH=/tools/openmpi/lib:$LD_LIBRARY_PATH
soapPath="/tools/SOAPdenovo2/SOAPdenovo-63mer"
workPath="/NGS"
outputPath="assembly/soap/demo"
/tools/openmpi/bin/mpirun $soapPath all -s $workPath/$outputPath/config_file -K 23 -R -F -p 60 -V -o $workPath/$outputPath/graph_prefix > $workPath/$outputPath/ass.log 2> $workPath/$outputPath/ass.err
2) Submit the job:
qsub -pe orte 60 mpi.qsub
3) The log in ass.err
a) It seemed it run soapdenovo several times according to the log
cat ass.err | grep "Pregraph" | wc -l
60
b) detail information
less ass.err (it seemed it only run soapdenov several times, because when I run it in my machine, it would only output one Pregraph):
Version 2.04: released on July 13th, 2012
Compile Apr 27 2016 15:50:02
********************
Pregraph
********************
Parameters: pregraph -s /NGS/assembly/soap/demo/config_file -K 23 -p 16 -R -o /NGS/assembly/soap/demo/graph_prefix
In /NGS/assembly/soap/demo/config_file, 1 lib(s), maximum read length 35, maximum name length 256.
Version 2.04: released on July 13th, 2012
Compile Apr 27 2016 15:50:02
********************
Pregraph
********************
and so on
c) information of stdin
cat ass.log:
--------------------------------------------------------------------------
WARNING: A process refused to die despite all the efforts!
This process may still be running and/or consuming resources.
Host: smp03
PID: 75035
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun noticed that process rank 58 with PID 0 on node c0214.local exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------
Information about cluster:
1) qconf -sql
all.q
smp.q
2) qconf -spl
mpi
mpich
orte
zhongxm
3) qconf -sp zhongxm
pe_name zhongxm
slots 999
user_lists NONE
xuser_lists NONE
start_proc_args /bin/true
stop_proc_args /bin/true
allocation_rule $fill_up
control_slaves TRUE
job_is_first_task FALSE
urgency_slots min
accounting_summary FALSE
4) qconf -sq smp.q
qname smp.q
hostlist #smp.q
seq_no 0
load_thresholds np_load_avg=1.75
suspend_thresholds NONE
nsuspend 1
suspend_interval 00:05:00
priority 0
min_cpu_interval 00:05:00
processors UNDEFINED
qtype BATCH INTERACTIVE
ckpt_list NONE
pe_list make zhongxm
rerun FALSE
slots 1
tmpdir /tmp
shell /bin/csh
prolog NONE
epilog NONE
shell_start_mode posix_compliant
starter_method NONE
suspend_method NONE
resume_method NONE
terminate_method NONE
notify 00:00:60
owner_list NONE
user_lists NONE
xuser_lists NONE
subordinate_list NONE
complex_values NONE
projects NONE
xprojects NONE
calendar NONE
initial_state default
s_rt INFINITY
h_rt INFINITY
s_cpu INFINITY
h_cpu INFINITY
s_fsize INFINITY
h_fsize INFINITY
s_data INFINITY
h_data INFINITY
s_stack INFINITY
h_stack INFINITY
s_core INFINITY
h_core INFINITY
s_rss INFINITY
h_rss INFINITY
s_vmem INFINITY
h_vmem INFINITY
5) qconf -sq all.q
qname all.q
hostlist #allhosts
seq_no 0
load_thresholds np_load_avg=1.75
suspend_thresholds NONE
nsuspend 1
suspend_interval 00:05:00
priority 0
min_cpu_interval 00:05:00
processors UNDEFINED
qtype BATCH INTERACTIVE
ckpt_list NONE
pe_list make zhongxm
rerun FALSE
slots 16,[c0219.local=32]
tmpdir /tmp
shell /bin/csh
prolog NONE
epilog NONE
shell_start_mode posix_compliant
starter_method NONE
suspend_method NONE
resume_method NONE
terminate_method NONE
notify 00:00:60
owner_list NONE
user_lists mobile
xuser_lists NONE
subordinate_list NONE
complex_values NONE
projects NONE
xprojects NONE
calendar NONE
initial_state default
s_rt INFINITY
h_rt INFINITY
s_cpu INFINITY
h_cpu INFINITY
s_fsize INFINITY
h_fsize INFINITY
s_data INFINITY
h_data INFINITY
s_stack INFINITY
h_stack INFINITY
s_core INFINITY
h_core INFINITY
s_rss INFINITY
h_rss INFINITY
s_vmem INFINITY
h_vmem INFINITY
According to https://hpc.unt.edu/soapdenovo the software doesn't support MPI:
This code is NOT compiled with MPI, and should only be used in parallel on a SINGLE node, via a threaded model.
So, you can't just start the software with mpiexec on cluster to have access to more memory. Cluster machines are connected with non-coherent networks (Ethernet, Infiniband) which are slower than memory bus, and PCs in cluster do not share their memory. Clusters use MPI libraries (OpenMPI or MPICH) to work with network, and all requests between nodes is explicit: program calls MPI_Send in one process and MPI_Recv in other. There are also one-way calls like MPI_Put/MPI_Get to access remote memory (RDMA - remote direct memory access), but this is not the same as local memory.
osgx, thank you for your reply very much and sorry for the delay of this message.
Since I don't major in computer, I think I can't understand some glossary very well, like ELF. So there are some new questions and I list my question as follow, thanks for help advace:
1) When I "ldd SOAPdenovo-63mer", it outputed "not a dynamic executable", did this mean "the code is not complied with MPI" that you mentioned?
2) In short, I can't solve the problem with the cluster, and I have to look for a machine with more than 512G memory?
3) Also, I used another software called ALLPATHS-LG (http://www.broadinstitute.org/software/allpaths-lg/blog/) that was also failed for shortage of memory, and according to FAQ C1 (http://www.broadinstitute.org/software/allpaths-lg/blog/?page_id=336), what "it uses share memory parallelization" mean, did it means it can use memory in cluster, or only memory in a node, and I have to find a machine with enough memory?
C1. Can I run ALLPATHS-LG on a cluster?
You can, but it will only use one machine, not the entire cluster. That machine would need to have enough memory to fit the entire assembly. ALLPATHS-LG does not support distributed computing using MPI, instead it uses Shared Memory Parallelization.
By the way, this is first time I posted here, I think I should use commit to reply, considering so many words, I use "Answer Your Question".

How to get number cores and processors form params

I'm new in shell scripts. I've got a sheduler, which start my script with qsub -l nodes=a:ppn=b myscript.sh. I need to get this a and b in myscript.sh. How can i get it?
Number of cores: e.g.
cat /proc/cpuinfo|grep -c processor
Not sure about the ppn nor about how to make a difference between cpus and cores though.
Okey, I've found the answer. In torque we have 2 variables for this. For get number of nodes u should use nnodes = $PBS_UM_NODES, for processors nprocs = $PBS_NUM_PPN

Linpack sometimes starting, sometimes not, but nothing changed

I installed Linpack on a 2-Node cluster with Xeon processors. Sometimes if I start Linpack with this command:
mpiexec -np 28 -print-rank-map -f /root/machines.HOSTS ./xhpl_intel64
linpack starts and prints the output, sometimes I only see the mpi mappings printed and then nothing following. To me this seems like random behaviour because I don't change anything between the calls and as already mentioned, Linpack sometimes starts, sometimes not.
In top I can see that xhpl_intel64processes have been created and they are heavily using the CPU but when watching the traffic between the nodes, iftop is telling me that it nothing is sent.
I am using MPICH2 as MPI implementation. This is my HPL.dat:
# cat HPL.dat
HPLinpack benchmark input file
Innovative Computing Laboratory, University of Tennessee
HPL.out output file name (if any)
6 device out (6=stdout,7=stderr,file)
1 # of problems sizes (N)
10000 Ns
1 # of NBs
250 NBs
0 PMAP process mapping (0=Row-,1=Column-major)
1 # of process grids (P x Q)
2 Ps
14 Qs
16.0 threshold
1 # of panel fact
2 PFACTs (0=left, 1=Crout, 2=Right)
1 # of recursive stopping criterium
4 NBMINs (>= 1)
1 # of panels in recursion
2 NDIVs
1 # of recursive panel fact.
1 RFACTs (0=left, 1=Crout, 2=Right)
1 # of broadcast
1 BCASTs (0=1rg,1=1rM,2=2rg,3=2rM,4=Lng,5=LnM)
1 # of lookahead depth
1 DEPTHs (>=0)
2 SWAP (0=bin-exch,1=long,2=mix)
64 swapping threshold
0 L1 in (0=transposed,1=no-transposed) form
0 U in (0=transposed,1=no-transposed) form
1 Equilibration (0=no,1=yes)
8 memory alignment in double (> 0)
edit2:
I now just let the program run for a while and after 30min it tells me:
# mpiexec -np 32 -print-rank-map -f /root/machines.HOSTS ./xhpl_intel64
(node-0:0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15)
(node-1:16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31)
Assertion failed in file ../../socksm.c at line 2577: (it_plfd->revents & 0x008) == 0
internal ABORT - process 0
APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)
Is this a mpi problem?
Do you know what type of problem this could be?
I figured out what the problem was: MPICH2 uses different random ports each time it starts and if these are blocked your application wont start up correctly.
The solution for MPICH2 is to set the environment variable MPICH_PORT_RANGE to START:END, like this:
export MPICH_PORT_RANGE=50000:51000
Best,
heinrich

Resources