Skip to content

feat: Mount DC in HPC sessions #1174

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sambuc wants to merge 1 commit into
base: sambuc/feat/upload-dcs-secrets
Choose a base branch
from
Open

feat: Mount DC in HPC sessions #1174

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
123 changes: 107 additions & 16 deletions internal/remote/firecrest/session_script.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,20 +36,49 @@ esac

: ${SESSION_DIR:="${PWD}"}
: ${SESSION_WORK_DIR:="${SESSION_DIR}/work"}
: ${SECRETS_DIR:="${SESSION_DIR}/secrets"}
: ${SECRETS_DIR:="$(mktemp -d)"}
: ${SECRETS_USER_DIR:="${SECRETS_DIR}/user/"}
: ${SECRETS_DATA_CONNECTORS_DIR:="${SECRETS_DIR}/data_connectors"}
: ${LOGS_DIR:="${SESSION_DIR}/logs"}
: ${CACHE_DIR:="${SECRETS_DIR}/cache"}

# Setup session environment
export RENKU_MOUNT_DIR="${SESSION_WORK_DIR}"
export RENKU_WORKING_DIR="${SESSION_WORK_DIR}"
# Force the frontend to listen on 127.0.0.1
export RENKU_SESSION_IP="127.0.0.1"

# Do not leave secrets on a shared fs, move it to the node where the session runs.
mkdir -p "${SECRETS_DIR}"
chmod 700 "${SECRETS_DIR}"
mv "${SESSION_DIR}/secrets"/* "${SECRETS_DIR}"
rmdir "${SESSION_DIR}/secrets"

Comment on lines +51 to +56

@leafty leafty Jul 17, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we do this? Official CSCS docs have example where user credentials are stored in the HOME folder. I really do not see what we gain by doing this.

Example: https://docs.cscs.ch/software/container-engine/run/#third-party-and-private-registries

Having the credentials in the HOME folder allows user to re-access the data connectors even outside of a remote Renku session (e.g. connecting directly to Alps with SSH).

# Load the wstunnel secret
export WSTUNNEL_SECRET="$(cat "${SECRETS_DIR}/wstunnel_secret")"

# CamelCase to kebab-case conversion
#
# Usage:
# kebab_cased="$(to_kebab_case "[cC]amelCaseInput")"
to_kebab_case() {
echo "${1:?"to_kebab_case: input string missing"}" | sed 's/\([A-Z]\)/-\1/g' | tr '[:upper:]' '[:lower:]'
}

# Convert camelCased key - value pairs stored in a flat json struct to command line arguments.
#
# WARNING: Whitespace and json reserved characters are stripped, so they cannot appear in keys nor values.
#
# Usage:
# arguments_string="$(to_rclone_mount_arguments "file_path" ["argument_prefix"])"
to_rclone_mount_arguments() {
local filename=${1:?"to_rclone_mount_arguments: input file missing"}
local prefix=${2}
cat "${filename}" | tr -d '{} \t\r\n' | sed -e 's/","/"\n"/g' -e 's/":"/": "/g' | while IFS=': ' read -r key value; do
printf "%s%s=%s " "${prefix}" "$(to_kebab_case "${key}")" "${value}"
done
}

# Installs rclone
#
# Usage:
Expand Down Expand Up @@ -135,8 +164,6 @@ for d in \
SESSION_DIR \
SESSION_WORK_DIR \
SECRETS_DIR \
SECRETS_USER_DIR \
SECRETS_DATA_CONNECTORS_DIR \
LOGS_DIR
do
echo "${d}: ${!d}"
Expand Down Expand Up @@ -172,20 +199,66 @@ fi
srun_param_container_image="--container-image ${REMOTE_SESSION_IMAGE}"
srun_param_workdir="--container-workdir ${SESSION_WORK_DIR}"
srun_param_mounts=#{{SESSION_MOUNTS_PLACEHOLDER}}
# We cannot generate directly the local secrets path from the proxy as the final path is only known after ${SECRETS_DIR} as been set.
srun_param_mounts=$(echo ${srun_param_mounts} | sed -e "s,${SESSION_DIR}/secrets,${SECRETS_DIR},g")

# Mount Data Connectors, if any
if [ -d "${SECRETS_DATA_CONNECTORS_DIR}" ]; then
(# Run in a sub shell to scope the temporary variables
for dc in "${SECRETS_DATA_CONNECTORS_DIR}"/*; do
n=$(echo ${dc}|sed -e 's,.*-,,')
mount="$(cat "${dc}/remote")"
mount_point="${SESSION_WORK_DIR}/${mount}"
remote_path="$(cat "${dc}/remotePath")"
log_file="${LOGS_DIR}/rclone-dc-${n}.log"
config_file="${dc}/configData"
pass="${dc}/pass"

if [ -f "${pass}" ]; then
pass_content="$(cat "${pass}" | ${rclone} obscure -)"
cat "${config_file}" | sed -e "s,pass *= <sensitive>,pass = ${pass_content}," > "${config_file}.tmp" && mv "${config_file}.tmp" "${config_file}"
rm -f "${pass}" || true
fi

echo "TODO: setup rclone mounts..."
if [ -f "${dc}/vfsOpt" ]; then
vfs_options="$(to_rclone_mount_arguments "${dc}/vfsOpt" "--vfs")"
fi

if [ -f "${dc}/mountOpt" ]; then
mount_options="$(to_rclone_mount_arguments "${dc}/mountOpt" "-")"
fi

# echo "Setting up example rclone mount..."
# fusermount3 -u "${SESSION_WORK_DIR}/era5" || true
# rm -rf "${SESSION_WORK_DIR}/era5"
# mkdir -p "${SESSION_WORK_DIR}/era5"
# RCLONE_CONFIG="${SESSION_DIR}/rclone.conf"
# cat <<EOF >"${RCLONE_CONFIG}"
# [era5]
# type = doi
# doi = 10.5281/zenodo.3831980
# EOF
# "${rclone}" mount --config "${RCLONE_CONFIG}" --daemon --read-only era5: "${SESSION_WORK_DIR}/era5"
if [ -f "${dc}/extraArgs" ]; then
extra_args="$(cat "${dc}/extraArgs")"
fi

echo >> "${log_file}"
echo "--- Starting $(date)" >> "${log_file}"

mkdir -p "${mount_point}"
mkdir -p "${CACHE_DIR}/${n}"

# Make sure there is no stale mount, and if so clean it up.
fusermount3 -uz "${mount_point}" 2>/dev/null || true
sleep 1 # Let the state stabilise before mounting something there

# We do our best to make sure we do not leave around bad rclone
# state, but sometimes there is still something lingering on...
# so we have to add --allow-non-empty.
Comment on lines +245 to +247

@leafty leafty Jul 17, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not do rm -rf <mount_point>; mkdir -p <mount_point>?

${rclone} mount \
--daemon \
--allow-non-empty \
${mount_options} \
--log-file="${log_file}" \
--cache-dir="${CACHE_DIR}/$n" \
${vfs_options} \
--config="${config_file}" \
${extra_args} \
"${mount}:${remote_path}" \
"//${mount_point}"
done
)
fi

# echo "Starting tunnel..."
echo "wstunnel client \
Expand Down Expand Up @@ -257,7 +330,25 @@ function exit_script() {
# Make sure we have a valid pid before attempting to kill it
(test -n "${pid}" && ps "${pid}" > /dev/null && kill -TERM "${pid}") || true

# fusermount3 -u "${SESSION_WORK_DIR}/era5" || true
# Cleanup Data Connector mount points
if [ -d "${SECRETS_DATA_CONNECTORS_DIR}" ]; then
for dc in "${SECRETS_DATA_CONNECTORS_DIR}"/*; do
fusermount3 -uz "${SESSION_WORK_DIR}/$(cat "${dc}/remote")" 2>/dev/null || true
rmdir "${SESSION_WORK_DIR}/$(cat "${dc}/remote")" || true
done
fi

# Remove the secrets from the node, leaving them generates problem in case
# of suppression of Data Connectors/user secrets from the project, for example.
# Same thing for the caches, as they will be renumbered.
for d in \
CACHE_DIR \
SECRETS_DATA_CONNECTORS_DIR \
SECRETS_USER_DIR \
SECRETS_DIR
do
rm -rf "${!d}" || true
done

# Sometimes the job continues to run...
scancel "${SLURM_JOB_ID}" || true
Expand Down
Loading