librelad c92fcc9773 fix(install): don't record an app as installed when its container never started
An image-pull failure (e.g. the MTU-EOF black hole) left apps marked
installed+active with no container: the compose `up -d` failure was invisible
because (a) checkSuccess read $? after `_rc=$?`, always printing "✓ Started",
and (b) the exit code is dropped across dockerComposeUpdateAndStartApp →
dockerComposeUpdate → dockerComposeRestartAfterUpdate, and installApp never
checked it — so post-install integrations ran and set status=1 regardless.

- up_app.sh: restore $? to the compose exit before checkSuccess (both rootless
  and rooted) so a failed `up -d` is reported as an error + logged, not "✓".
- app_install.sh: after `up`, reality-gate on the app's compose project having
  at least one container (ps -a, so a slow-to-start container still counts).
  If none exists, print a clear failure, skip _appPostStartIntegrations (which
  is what records the app + sets status=1), and return non-zero.

Verified: the gate query passes for a running app (libreportal) and refuses
apps with no container (navidrome/ipinfo after their pulls EOF'd).

Signed-off-by: librelad <librelad@digitalangels.vip>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-06 20:43:56 +01:00

154 lines
8.0 KiB
Bash
Executable File

#!/bin/bash
dockerComposeUp()
{
local app_name="$1"
local custom_compose="$2"
local type="$3"
if [[ "$app_name" == "" ]]; then
isError "Something went wrong...No app name provided..."
return 1
fi
# Real exit status of the compose command, so callers that gate on it (the
# updater's rollback, the artifact apply pipeline) aren't fail-open. checkSuccess
# only logs; it does NOT set the function's return — without this the function
# always returned 0 and a failed `up -d` looked like success.
local _rc=0
isHeader "Docker Compose Up $app_name"
# Make sure we are able to get the compose file
if [[ $compose_setup == "" ]]; then
setupBasicScanVariables "$app_name"
fi
# Compose file public variable for restarting etc
if [[ $compose_setup == "default" ]]; then
local setup_compose="-f docker-compose.yml"
local compose_file="docker-compose.yml"
elif [[ $compose_setup == "app" ]]; then
local setup_compose="-f docker-compose.yml -f docker-compose.$app_name.yml"
local compose_file="docker-compose.$app_name.yml"
fi
if [[ $custom_compose != "" ]]; then
local setup_compose="-f docker-compose.yml -f $custom_compose"
local compose_file="$custom_compose"
fi
if [[ "$OS_TYPE" == "Ubuntu" || "$OS_TYPE" == "Debian" ]]; then
if [ -f "$containers_dir$app_name/$compose_file" ]; then
# Quiet pull + plain progress so progress redraws don't flood the log.
local _compose_quiet="--quiet-pull"
export COMPOSE_PROGRESS=plain
# Force a rebuild when the app ships a Dockerfile. Without
# --build, `up -d` reuses the cached image and silently
# ignores any edits to Dockerfile / source between installs.
local _compose_build_flag=""
local _is_local_build=0
if [[ -f "$containers_dir$app_name/Dockerfile" ]]; then
_compose_build_flag="--build"
_is_local_build=1
fi
# Used for the standard LibrePortal app
if [[ "$type" == "" ]]; then
# Refuse to start if the runtime compose still has raw
# `#LIBREPORTAL|TAG|VALUE` placeholders on the value side.
# Happens when the template was copied over the runtime
# compose without the tag processors running (manual
# copy, partial restore, aborted install). Without this
# guard docker errors with "invalid boolean:
# HEALTHCHECK_DATA" or similar.
#
# Two-pass: detect → attempt self-heal by re-running the
# full tag-processor pipeline → re-detect. Only refuse
# if heal didn't clear all placeholders.
_scanStaleTags() {
awk '
{
idx = index($0, "#LIBREPORTAL|")
if (idx == 0) next
# Skip whole-line YAML comments (e.g. "# - PORTS_DATA_4 ...")
value_part = substr($0, 1, idx - 1)
trimmed = value_part
sub(/^[ \t]+/, "", trimmed)
if (substr(trimmed, 1, 1) == "#") next
marker = substr($0, idx + length("#LIBREPORTAL|"))
if (split(marker, parts, "|") < 2) next
placeholder = parts[2]
gsub(/[ \t]+$/, "", placeholder)
if (placeholder ~ /^[A-Z][A-Z0-9_]*_DATA(_[0-9]+)?$/) {
printf " line %d: %s\n", NR, $0
}
}
' "$1"
}
local _compose_path="$containers_dir$app_name/$compose_file"
local _stale_tags
_stale_tags=$(_scanStaleTags "$_compose_path")
if [[ -n "$_stale_tags" ]]; then
isNotice "$app_name compose has unsubstituted LibrePortal tag placeholders — attempting self-heal via tag processors."
if declare -F dockerConfigSetupFileWithData >/dev/null 2>&1; then
# The processors read per-app env vars
# (healthcheck, host_setup, app_category, …) set
# by initializeAppVariables. Without it some
# tags get filled with empty strings and the
# heal leaves placeholders behind.
if declare -F initializeAppVariables >/dev/null 2>&1; then
initializeAppVariables "$app_name" >/dev/null 2>&1 || true
fi
dockerConfigSetupFileWithData "$app_name" >/dev/null 2>&1 || true
_stale_tags=$(_scanStaleTags "$_compose_path")
fi
if [[ -n "$_stale_tags" ]]; then
isError "$app_name compose still has unsubstituted LibrePortal tag placeholders after self-heal — refusing to start."
isNotice "File: $_compose_path"
while IFS= read -r _l; do isNotice "$_l"; done <<< "$_stale_tags"
isNotice "Fix: run 'libreportal app install $app_name' to re-apply the full install pipeline."
unset -f _scanStaleTags
return 1
fi
isSuccessful "Tag processors re-applied — placeholders resolved, continuing start."
fi
unset -f _scanStaleTags
# Local-build apps take noticeably longer than image-pull
# apps. Specific heads-up so the user doesn't wonder if
# it's stuck during a multi-minute Dockerfile build.
if (( _is_local_build )); then
isNotice "$app_name has a Dockerfile — building image locally."
isNotice "First build can take a few minutes."
fi
if [[ $CFG_DOCKER_INSTALL_TYPE == "rootless" ]]; then
isNotice "Starting container for $app_name, this may take a while..."
local result; result=$(dockerCommandRunInstallUser "cd $containers_dir$app_name && COMPOSE_PROGRESS=plain docker compose $setup_compose up $_compose_quiet $_compose_build_flag -d")
_rc=$?
# Restore $? to the compose exit code — a bare `checkSuccess`
# after `_rc=$?` would read the assignment's 0 and always print
# success, hiding a failed `up -d` (e.g. an image pull EOF).
( exit "$_rc" ); checkSuccess "Started container for $app_name"
elif [[ $CFG_DOCKER_INSTALL_TYPE == "rooted" ]]; then
isNotice "Starting container for $app_name, this may take a while..."
local result; result=$(cd "$containers_dir$app_name" && COMPOSE_PROGRESS=plain docker compose $setup_compose up $_compose_quiet $_compose_build_flag -d)
_rc=$?
( exit "$_rc" ); checkSuccess "Started container for $app_name"
fi
# Used for the CLI dockertype switcher.
else
if [[ $type == "rootless" ]]; then
local result; result=$(dockerCommandRunInstallUser "cd $containers_dir$app_name && docker compose $setup_compose down"); _rc=$?
checkSuccess "Shutting down container for $app_name"
elif [[ $type == "rooted" ]]; then
local result; result=$(cd "$containers_dir$app_name" && docker compose $setup_compose down); _rc=$?
checkSuccess "Shutting down container for $app_name"
fi
fi
else
isNotice "Unable to find the compose file to docker compose up this application."
_rc=1
fi
fi
return $_rc
}