forked from rarias/jungle
Compare commits
7 Commits
fix/cross-
...
100e4e7029
| Author | SHA1 | Date | |
|---|---|---|---|
|
100e4e7029
|
|||
|
8d56802b1c
|
|||
|
bcadb12ce2
|
|||
|
7268da81d2
|
|||
|
1b7d3a6e8e
|
|||
|
0b6af59af8
|
|||
|
092326c4e6
|
@@ -93,4 +93,20 @@
|
|||||||
wantedBy = [ "multi-user.target" ];
|
wantedBy = [ "multi-user.target" ];
|
||||||
serviceConfig.ExecStart = script;
|
serviceConfig.ExecStart = script;
|
||||||
};
|
};
|
||||||
|
|
||||||
|
# Only allow SSH connections from users who have a SLURM allocation
|
||||||
|
# See: https://slurm.schedmd.com/pam_slurm_adopt.html
|
||||||
|
security.pam.services.sshd.rules.account.slurm = {
|
||||||
|
control = "required";
|
||||||
|
enable = true;
|
||||||
|
modulePath = "${pkgs.slurm}/lib/security/pam_slurm_adopt.so";
|
||||||
|
args = [ "log_level=debug5" ];
|
||||||
|
order = 999999; # Make it last one
|
||||||
|
};
|
||||||
|
|
||||||
|
# Disable systemd session (pam_systemd.so) as it will conflict with the
|
||||||
|
# pam_slurm_adopt.so module. What happens is that the shell is first adopted
|
||||||
|
# into the slurmstepd task and then into the systemd session, which is not
|
||||||
|
# what we want, otherwise it will linger even if all jobs are gone.
|
||||||
|
security.pam.services.sshd.startSession = lib.mkForce false;
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -1,4 +1,4 @@
|
|||||||
{ lib, pkgs, ... }:
|
{ lib, ... }:
|
||||||
|
|
||||||
{
|
{
|
||||||
imports = [
|
imports = [
|
||||||
@@ -21,20 +21,4 @@
|
|||||||
};
|
};
|
||||||
|
|
||||||
services.slurm.client.enable = true;
|
services.slurm.client.enable = true;
|
||||||
|
|
||||||
# Only allow SSH connections from users who have a SLURM allocation
|
|
||||||
# See: https://slurm.schedmd.com/pam_slurm_adopt.html
|
|
||||||
security.pam.services.sshd.rules.account.slurm = {
|
|
||||||
control = "required";
|
|
||||||
enable = true;
|
|
||||||
modulePath = "${pkgs.slurm}/lib/security/pam_slurm_adopt.so";
|
|
||||||
args = [ "log_level=debug5" ];
|
|
||||||
order = 999999; # Make it last one
|
|
||||||
};
|
|
||||||
|
|
||||||
# Disable systemd session (pam_systemd.so) as it will conflict with the
|
|
||||||
# pam_slurm_adopt.so module. What happens is that the shell is first adopted
|
|
||||||
# into the slurmstepd task and then into the systemd session, which is not
|
|
||||||
# what we want, otherwise it will linger even if all jobs are gone.
|
|
||||||
security.pam.services.sshd.startSession = lib.mkForce false;
|
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -1,6 +1,31 @@
|
|||||||
{ config, pkgs, ... }:
|
{ config, pkgs, ... }:
|
||||||
|
|
||||||
{
|
let
|
||||||
|
suspendProgram = pkgs.writeShellScript "suspend.sh" ''
|
||||||
|
exec 1>>/var/log/power_save.log 2>>/var/log/power_save.log
|
||||||
|
set -x
|
||||||
|
export "PATH=/run/current-system/sw/bin:$PATH"
|
||||||
|
echo "$(date) Suspend invoked $0 $*" >> /var/log/power_save.log
|
||||||
|
hosts=$(scontrol show hostnames $1)
|
||||||
|
for host in $hosts; do
|
||||||
|
echo Shutting down host: $host
|
||||||
|
ipmitool -I lanplus -H ''${host}-ipmi -P "" -U "" chassis power off
|
||||||
|
done
|
||||||
|
'';
|
||||||
|
|
||||||
|
resumeProgram = pkgs.writeShellScript "resume.sh" ''
|
||||||
|
exec 1>>/var/log/power_save.log 2>>/var/log/power_save.log
|
||||||
|
set -x
|
||||||
|
export "PATH=/run/current-system/sw/bin:$PATH"
|
||||||
|
echo "$(date) Suspend invoked $0 $*" >> /var/log/power_save.log
|
||||||
|
hosts=$(scontrol show hostnames $1)
|
||||||
|
for host in $hosts; do
|
||||||
|
echo Starting host: $host
|
||||||
|
ipmitool -I lanplus -H ''${host}-ipmi -P "" -U "" chassis power on
|
||||||
|
done
|
||||||
|
'';
|
||||||
|
|
||||||
|
in {
|
||||||
services.slurm = {
|
services.slurm = {
|
||||||
controlMachine = "apex";
|
controlMachine = "apex";
|
||||||
clusterName = "jungle";
|
clusterName = "jungle";
|
||||||
@@ -34,6 +59,16 @@
|
|||||||
# the resources. Use the task/cgroup plugin to enable process containment.
|
# the resources. Use the task/cgroup plugin to enable process containment.
|
||||||
TaskPlugin=task/affinity,task/cgroup
|
TaskPlugin=task/affinity,task/cgroup
|
||||||
|
|
||||||
|
# Power off unused nodes until they are requested
|
||||||
|
SuspendProgram=${suspendProgram}
|
||||||
|
SuspendTimeout=60
|
||||||
|
ResumeProgram=${resumeProgram}
|
||||||
|
ResumeTimeout=300
|
||||||
|
SuspendExcNodes=fox
|
||||||
|
|
||||||
|
# Turn the nodes off after 1 hour of inactivity
|
||||||
|
SuspendTime=3600
|
||||||
|
|
||||||
# Reduce port range so we can allow only this range in the firewall
|
# Reduce port range so we can allow only this range in the firewall
|
||||||
SrunPortRange=60000-61000
|
SrunPortRange=60000-61000
|
||||||
|
|
||||||
|
|||||||
@@ -43,7 +43,7 @@ stdenv.mkDerivation rec {
|
|||||||
|
|
||||||
configureFlags = [
|
configureFlags = [
|
||||||
"--with-infiniband=${rdma-core-all}"
|
"--with-infiniband=${rdma-core-all}"
|
||||||
"--with-mpi=yes" # fixes mpi detection when cross-compiling
|
"--with-mpi=yes"
|
||||||
"--with-slurm"
|
"--with-slurm"
|
||||||
"CFLAGS=-fPIC"
|
"CFLAGS=-fPIC"
|
||||||
"CXXFLAGS=-fPIC"
|
"CXXFLAGS=-fPIC"
|
||||||
@@ -70,6 +70,6 @@ stdenv.mkDerivation rec {
|
|||||||
maintainers = with lib.maintainers.bsc; [ rarias ];
|
maintainers = with lib.maintainers.bsc; [ rarias ];
|
||||||
platforms = lib.platforms.linux;
|
platforms = lib.platforms.linux;
|
||||||
license = lib.licenses.gpl3Plus;
|
license = lib.licenses.gpl3Plus;
|
||||||
cross = false; # infiniband detection does not work
|
cross = false;
|
||||||
};
|
};
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -17,7 +17,6 @@
|
|||||||
, gitUrl ? "ssh://git@bscpm04.bsc.es/llvm-ompss/llvm-mono.git"
|
, gitUrl ? "ssh://git@bscpm04.bsc.es/llvm-ompss/llvm-mono.git"
|
||||||
, gitBranch ? "master"
|
, gitBranch ? "master"
|
||||||
, gitCommit ? "872ba63f86edaefc9787984ef3fae9f2f94e0124" # github-release-2025.11
|
, gitCommit ? "872ba63f86edaefc9787984ef3fae9f2f94e0124" # github-release-2025.11
|
||||||
, buildLlvmPackages
|
|
||||||
}:
|
}:
|
||||||
|
|
||||||
let
|
let
|
||||||
@@ -120,8 +119,6 @@ in stdenv.mkDerivation {
|
|||||||
"-DCMAKE_INSTALL_RPATH=${zlib}/lib:${gcc.cc.lib}/lib"
|
"-DCMAKE_INSTALL_RPATH=${zlib}/lib:${gcc.cc.lib}/lib"
|
||||||
"-DLLVM_APPEND_VC_REV=ON"
|
"-DLLVM_APPEND_VC_REV=ON"
|
||||||
"-DLLVM_FORCE_VC_REVISION=${source.version}"
|
"-DLLVM_FORCE_VC_REVISION=${source.version}"
|
||||||
"-DLLVM_TABLEGEN_EXE=${buildLlvmPackages.tblgen}/bin/llvm-tblgen"
|
|
||||||
"-DCLANG_TABLEGEN=${buildLlvmPackages.tblgen}/bin/clang-tblgen"
|
|
||||||
)
|
)
|
||||||
'';
|
'';
|
||||||
|
|
||||||
|
|||||||
@@ -47,7 +47,7 @@ in mpich.overrideAttrs (old: {
|
|||||||
|
|
||||||
preFixup = ''
|
preFixup = ''
|
||||||
sed -i 's:^CC=.*:CC=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}cc:' $out/bin/mpicc
|
sed -i 's:^CC=.*:CC=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}cc:' $out/bin/mpicc
|
||||||
sed -i 's:^CXX=.*:CXX=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}c++:' $out/bin/mpicxx
|
sed -i 's:^CXX=.*:CXX=${targetPackages.stdenv.cc}/bin/${targetPackages.stdenv.cc.targetPrefix}++:' $out/bin/mpicxx
|
||||||
'' + lib.optionalString enableFortran ''
|
'' + lib.optionalString enableFortran ''
|
||||||
sed -i 's:^FC=.*:FC=${targetPackages.gfortran or gfortran}/bin/${targetPackages.gfortran.targetPrefix or gfortran.targetPrefix}gfortran:' $out/bin/mpifort
|
sed -i 's:^FC=.*:FC=${targetPackages.gfortran or gfortran}/bin/${targetPackages.gfortran.targetPrefix or gfortran.targetPrefix}gfortran:' $out/bin/mpifort
|
||||||
'';
|
'';
|
||||||
|
|||||||
@@ -34,7 +34,7 @@ stdenv.mkDerivation rec {
|
|||||||
|
|
||||||
env = {
|
env = {
|
||||||
MPICH_CC="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}cc";
|
MPICH_CC="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}cc";
|
||||||
MPICH_CXX="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}c++";
|
MPICH_CXX="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}++";
|
||||||
};
|
};
|
||||||
|
|
||||||
postInstall = ''
|
postInstall = ''
|
||||||
|
|||||||
Reference in New Issue
Block a user