Merge pull request #437 from boegel/2023.06-software.eessi.io_link_nv…

…idia_host_libraries_fix make `link_nvidia_host_libraries.sh` script a bit more robust, in case target of host_injections directory is a non-existing directory
EESSI · Dec 22, 2023 · 44b563c · 44b563c
2 parents 5c322b0 + ac53cf0
commit 44b563c
Show file tree

Hide file tree

Showing 2 changed files with 52 additions and 39 deletions.
diff --git a/EESSI-install-software.sh b/EESSI-install-software.sh
@@ -203,42 +203,47 @@ ${EESSI_PREFIX}/scripts/gpu_support/nvidia/install_cuda_host_injections.sh -c 12
 # ${EESSI_PREFIX}/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh
 
 # use PR patch file to determine in which easystack files stuff was added
-for easystack_file in $(cat ${pr_diff} | grep '^+++' | cut -f2 -d' ' | sed 's@^[a-z]/@@g' | grep '^easystacks/.*yml$' | egrep -v 'known-issues|missing'); do
-
-    echo -e "Processing easystack file ${easystack_file}...\n\n"
-
-    # determine version of EasyBuild module to load based on EasyBuild version included in name of easystack file
-    eb_version=$(echo ${easystack_file} | sed 's/.*eb-\([0-9.]*\).*/\1/g')
-
-    # load EasyBuild module (will be installed if it's not available yet)
-    source ${TOPDIR}/load_easybuild_module.sh ${eb_version}
-
-    ${EB} --show-config
-
-    echo_green "All set, let's start installing some software with EasyBuild v${eb_version} in ${EASYBUILD_INSTALLPATH}..."
-
-    if [ -f ${easystack_file} ]; then
-        echo_green "Feeding easystack file ${easystack_file} to EasyBuild..."
-
-        ${EB} --easystack ${TOPDIR}/${easystack_file} --robot
-        ec=$?
-
-        # copy EasyBuild log file if EasyBuild exited with an error
-        if [ ${ec} -ne 0 ]; then
-            eb_last_log=$(unset EB_VERBOSE; eb --last-log)
-            # copy to current working directory
-            cp -a ${eb_last_log} .
-            echo "Last EasyBuild log file copied from ${eb_last_log} to ${PWD}"
-            # copy to build logs dir (with context added)
-            copy_build_log "${eb_last_log}" "${build_logs_dir}"
+changed_easystacks=$(cat ${pr_diff} | grep '^+++' | cut -f2 -d' ' | sed 's@^[a-z]/@@g' | grep '^easystacks/.*yml$' | egrep -v 'known-issues|missing') 
+if [ -z ${changed_easystacks} ]; then
+    echo "No missing installations, party time!"  # Ensure the bot report success, as there was nothing to be build here
+else
+    for easystack_file in ${changed_easystacks}; do
+
+        echo -e "Processing easystack file ${easystack_file}...\n\n"
+
+        # determine version of EasyBuild module to load based on EasyBuild version included in name of easystack file
+        eb_version=$(echo ${easystack_file} | sed 's/.*eb-\([0-9.]*\).*/\1/g')
+
+        # load EasyBuild module (will be installed if it's not available yet)
+        source ${TOPDIR}/load_easybuild_module.sh ${eb_version}
+
+        ${EB} --show-config
+
+        echo_green "All set, let's start installing some software with EasyBuild v${eb_version} in ${EASYBUILD_INSTALLPATH}..."
+
+        if [ -f ${easystack_file} ]; then
+            echo_green "Feeding easystack file ${easystack_file} to EasyBuild..."
+
+            ${EB} --easystack ${TOPDIR}/${easystack_file} --robot
+            ec=$?
+
+            # copy EasyBuild log file if EasyBuild exited with an error
+            if [ ${ec} -ne 0 ]; then
+                eb_last_log=$(unset EB_VERBOSE; eb --last-log)
+                # copy to current working directory
+                cp -a ${eb_last_log} .
+                echo "Last EasyBuild log file copied from ${eb_last_log} to ${PWD}"
+                # copy to build logs dir (with context added)
+                copy_build_log "${eb_last_log}" "${build_logs_dir}"
+            fi
+
+            $TOPDIR/check_missing_installations.sh ${TOPDIR}/${easystack_file}
+        else
+            fatal_error "Easystack file ${easystack_file} not found!"
         fi
-
-        $TOPDIR/check_missing_installations.sh ${TOPDIR}/${easystack_file}
-    else
-        fatal_error "Easystack file ${easystack_file} not found!"
-    fi
-
-done
+
+    done
+fi
 
 ### add packages here
 

diff --git a/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh b/scripts/gpu_support/nvidia/link_nvidia_host_libraries.sh
@@ -48,8 +48,10 @@ export LD_LIBRARY_PATH=/.singularity.d/libs:$LD_LIBRARY_PATH
 nvidia_smi_command="nvidia-smi --query-gpu=driver_version --format=csv,noheader"
 if $nvidia_smi_command > /dev/null; then
   host_driver_version=$($nvidia_smi_command | tail -n1)
+  echo_green "Found NVIDIA GPU driver version ${host_driver_version}"
   # If the first worked, this should work too
   host_cuda_version=$(nvidia-smi  -q --display=COMPUTE | grep CUDA | awk 'NF>1{print $NF}')
+  echo_green "Found host CUDA version ${host_cuda_version}"
 else
   error="Failed to successfully execute\n  $nvidia_smi_command\n"
   fatal_error "$error"
@@ -58,12 +60,18 @@ fi
 # Let's make sure the driver libraries are not already in place
 link_drivers=1
 
+# first make sure that target of host_injections variant symlink is an existing directory
+host_injections_target=$(realpath -m ${EESSI_CVMFS_REPO}/host_injections)
+if [ ! -d ${host_injections_target} ]; then
+    create_directory_structure ${host_injections_target}
+fi
+
 host_injections_nvidia_dir="${EESSI_CVMFS_REPO}/host_injections/nvidia/${EESSI_CPU_FAMILY}"
 host_injection_driver_dir="${host_injections_nvidia_dir}/host"
 host_injection_driver_version_file="$host_injection_driver_dir/driver_version.txt"
 if [ -e "$host_injection_driver_version_file" ]; then
   if grep -q "$host_driver_version" "$host_injection_driver_version_file"; then
-    echo_green "The host CUDA driver libraries have already been linked!"
+    echo_green "The host GPU driver libraries (v${host_driver_version}) have already been linked! (based on ${host_injection_driver_version_file})"
     link_drivers=0
   else
     # There's something there but it is out of date
@@ -91,8 +99,8 @@ if [ "$link_drivers" -eq 1 ]; then
   ls /.singularity.d/libs/* >> "$temp_dir"/libs.txt 2>/dev/null
 
   # Leverage singularity to find the full list of libraries we should be linking to
-  echo_yellow "Downloading latest version of nvliblist.conf from Apptainer"
-  curl -o "$temp_dir"/nvliblist.conf https://raw.githubusercontent.com/apptainer/apptainer/main/etc/nvliblist.conf
+  echo_yellow "Downloading latest version of nvliblist.conf from Apptainer to ${temp_dir}/nvliblist.conf"
+  curl --silent --output "$temp_dir"/nvliblist.conf https://raw.githubusercontent.com/apptainer/apptainer/main/etc/nvliblist.conf
 
   # Make symlinks to all the interesting libraries
   grep '.so$' "$temp_dir"/nvliblist.conf | xargs -i grep {} "$temp_dir"/libs.txt | xargs -i ln -s {}
@@ -133,4 +141,4 @@ else
   ln -s $host_injections_nvidia_dir/latest lib
 fi
 
-echo_green "Host NVIDIA gpu drivers linked successfully for EESSI"
+echo_green "Host NVIDIA GPU drivers linked successfully for EESSI"