Compare commits

..

5 Commits

10 changed files with 65 additions and 47 deletions

View File

@@ -93,4 +93,20 @@
wantedBy = [ "multi-user.target" ];
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;
}

View File

@@ -1,4 +1,4 @@
{ lib, pkgs, ... }:
{ lib, ... }:
{
imports = [
@@ -21,20 +21,4 @@
};
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;
}

View File

@@ -1,6 +1,31 @@
{ 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 = {
controlMachine = "apex";
clusterName = "jungle";
@@ -34,6 +59,16 @@
# the resources. Use the task/cgroup plugin to enable process containment.
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
SrunPortRange=60000-61000

View File

@@ -101,16 +101,14 @@ let
pkgsTopLevel = filterAttrs (_: isDerivation) bscPkgs;
# Native build in that platform doesn't imply cross build works
canCrossCompile = platform: default: pkg:
canCrossCompile = platform: pkg:
(isDerivation pkg) &&
# If meta.cross is undefined, use default
(pkg.meta.cross or default) &&
(meta.availableOn final.pkgsCross.${platform}.stdenv.hostPlatform pkg);
# Must be defined explicitly
(pkg.meta.cross or false) &&
(meta.availableOn platform pkg);
# For now only RISC-V
crossSet = genAttrs [ "riscv64" ] (platform:
filterAttrs (_: canCrossCompile platform true)
final.pkgsCross.${platform}.bsc.pkgsTopLevel);
crossSet = { riscv64 = final.pkgsCross.riscv64.bsc.pkgsTopLevel; };
buildList = name: paths:
final.runCommandLocal name { } ''
@@ -130,7 +128,7 @@ let
# For now only RISC-V
crossList = buildList "ci-cross"
(filter
(canCrossCompile "riscv64" false) # opt-in (pkgs with: meta.cross = true)
(canCrossCompile final.pkgsCross.riscv64.stdenv.hostPlatform)
(builtins.attrValues crossSet.riscv64));
in bscPkgs // {

View File

@@ -90,7 +90,7 @@ in
meta = {
description = "Performance analysis tool-suite for x86 based applications";
homepage = "https://www.amd.com/es/developer/uprof.html";
platforms = [ "x86_64-linux" ];
platforms = lib.platforms.linux;
license = lib.licenses.unfree;
maintainers = with lib.maintainers.bsc; [ rarias varcila ];
};

View File

@@ -1,6 +1,5 @@
{
stdenv
, lib
, cudatoolkit
, cudaPackages
, autoAddDriverRunpath
@@ -41,9 +40,4 @@ stdenv.mkDerivation (finalAttrs: {
'';
installPhase = "touch $out";
};
meta = {
platforms = [ "x86_64-linux" ];
maintainers = with lib.maintainers.bsc; [ rarias ];
};
})

View File

@@ -43,7 +43,7 @@ stdenv.mkDerivation rec {
configureFlags = [
"--with-infiniband=${rdma-core-all}"
"--with-mpi=yes" # fixes mpi detection when cross-compiling
"--with-mpi=yes"
"--with-slurm"
"CFLAGS=-fPIC"
"CXXFLAGS=-fPIC"
@@ -70,6 +70,6 @@ stdenv.mkDerivation rec {
maintainers = with lib.maintainers.bsc; [ rarias ];
platforms = lib.platforms.linux;
license = lib.licenses.gpl3Plus;
cross = false; # infiniband detection does not work
cross = false;
};
}

View File

@@ -17,7 +17,6 @@
, gitUrl ? "ssh://git@bscpm04.bsc.es/llvm-ompss/llvm-mono.git"
, gitBranch ? "master"
, gitCommit ? "872ba63f86edaefc9787984ef3fae9f2f94e0124" # github-release-2025.11
, buildLlvmPackages
}:
let
@@ -120,8 +119,6 @@ in stdenv.mkDerivation {
"-DCMAKE_INSTALL_RPATH=${zlib}/lib:${gcc.cc.lib}/lib"
"-DLLVM_APPEND_VC_REV=ON"
"-DLLVM_FORCE_VC_REVISION=${source.version}"
"-DLLVM_TABLEGEN_EXE=${buildLlvmPackages.tblgen}/bin/llvm-tblgen"
"-DCLANG_TABLEGEN=${buildLlvmPackages.tblgen}/bin/clang-tblgen"
)
'';

View File

@@ -12,7 +12,6 @@
# For each arch
, enableFortran ? stdenv.hostPlatform == stdenv.buildPlatform
, perl
, targetPackages
}:
let
@@ -46,10 +45,10 @@ in mpich.overrideAttrs (old: {
];
preFixup = ''
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++:' $out/bin/mpicxx
sed -i "s:^CC=.*:CC=''${CC:-gcc}:" $out/bin/mpicc
sed -i "s:^CXX=.*:CXX=''${CXX:-g++}:" $out/bin/mpicxx
'' + lib.optionalString enableFortran ''
sed -i 's:^FC=.*:FC=${targetPackages.gfortran or gfortran}/bin/${targetPackages.gfortran.targetPrefix or gfortran.targetPrefix}gfortran:' $out/bin/mpifort
sed -i "s:^FC=.*:FC=''${FC:-gfortran}:" $out/bin/mpifort
'';
hardeningDisable = [ "all" ];

View File

@@ -32,11 +32,6 @@ stdenv.mkDerivation rec {
"CXX=mpicxx"
];
env = {
MPICH_CC="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}cc";
MPICH_CXX="${stdenv.cc}/bin/${stdenv.cc.targetPrefix}c++";
};
postInstall = ''
mkdir -p $out/bin
for f in $(find $out -executable -type f); do