Add apt_update_safe() function that warns instead of aborting when apt-get update fails (e.g. enterprise repo 401 Unauthorized). Shows a helpful hint about disabling the enterprise repo when no subscription is active. Replaces direct apt-get update calls in build.func and install.func.
Detect Alpine inside the LXC container and install Tailscale via apk (add community repo if missing), enable/start the service. Preserve Debian/Ubuntu install path but improve DNS resolution checks, temporarily override /etc/resolv.conf if DNS appears blocked, and restore it afterwards. Switch pct exec to use sh -c, tighten command existence checks and redirections, ensure curl and keyring directory are present, add Tailscale apt source and install package. Overall robustness and error-handling improvements for installing Tailscale in containers.
When extracting GitHub source URLs in the workflow, only search the "# Source:" line first to avoid matching other URLs (such as license links). Update the grep pipeline to filter for the Source line (case-insensitive) before extracting the https://github.com/... pattern and add explanatory comments.
Prevent terminal job-control signals from suspending the script during recovery by trapping TSTP, TTIN and TTOU (instead of only TSTP) and restoring them on exit. Also clear the terminal 'tostop' flag in stop_spinner() with `stty -tostop` to avoid background spinner I/O from stopping the process group.
* fix(zammad): configure Elasticsearch for LXC container startup
- Set discovery.type: single-node (required for single-node ES)
- Set xpack.security.enabled: false (not needed in local LXC)
- Set bootstrap.memory_lock: false (fails in unprivileged LXC)
- Add startup wait loop (up to 60s) to ensure ES is ready before
Zammad installation continues
Fixes #12301-related recurring Elasticsearch startup failures
* refactor(api): eliminate duplicate traps, harden error handling & telemetry
Phase 1 - Structural:
- Remove api_exit_script() and 5 inline traps from build.func
- error_handler.func is now the sole trap owner via catch_errors()
- Update api.func comment reference (api_exit_script -> on_exit)
Phase 2 - Quality:
- Add stop_spinner() + cursor restore to error_handler(), on_interrupt(),
on_terminate(), on_hangup() to prevent spinner/cursor artifacts
- Enhance _send_abort_telemetry() with error text (last 20 log lines),
duration calculation, and 2 retry attempts (was fire-and-forget)
- Harden json_escape() to also strip DEL (0x7F) character
* fix(build): show spinner during post_update_to_api to prevent Ctrl+Z abort
post_update_to_api can take up to 33 seconds worst-case (3 curl attempts
x 10s timeout + sleep delays). Without any terminal output during this
time, users think the script is stuck and press Ctrl+Z, which prevents
the recovery menu from ever appearing.
Add msg_info spinner before both post_update_to_api calls in the failure
path (initial report + final force retry after recovery menu).
* fix(build): prevent SIGTSTP from killing recovery dialog
- Replace msg_info/stop_spinner with plain echo for telemetry reporting
The background spinner process in non-interactive shells (bash -c)
can trigger SIGTSTP, stopping the entire process group before the
recovery dialog appears. Plain echo avoids this.
- Add trap '' TSTP at failure path entry to ignore suspension signals
Prevents Ctrl+Z or terminal-related SIGTSTP from interrupting the
recovery menu. Restored with trap - TSTP before exit.
- Root cause: msg_info starts a background process (spinner &) that
is not properly detached in non-interactive shells where job control
(set -m) is OFF. The disown builtin has no effect without job
control, leaving the spinner in the same process group. This can
cause terminal I/O conflicts during the 33-second post_update_to_api
retry window, resulting in [2]+ Stopped.
* fix(test): initialize colors and remove illegal local in test harness
- Call load_functions() after sourcing core.func to initialize
color/formatting/icon variables (RD, GN, YW, CL, TAB, etc.)
- Remove 'local' keyword from top-level scope (not inside function)
- Default REPO_SOURCE to ref_api instead of main
* chore: remove test-recovery-dialog.sh from branch
* Revert "fix(zammad): configure Elasticsearch for LXC container startup"
This reverts commit 10e450b72f.
* fix(build): show telemetry status only in verbose mode
Telemetry reporting is an implementation detail that doesn't help
the user during failure recovery. Wrap echo statements with
VERBOSE check so they only appear when verbose mode is enabled.
* fix(workflow): improve Node.js version drift detection accuracy
1. Fix source URL regex: now captures 'Github: https://github.com/...'
pattern (pipe-separated), not just '# Source: https://github.com/...'
This was causing ~50 scripts to show 'No GitHub source'
2. Fix semver comparison: engines.node constraints like '>=18.0.0'
no longer flag version 22 as drift. >= and ^ constraints are now
properly evaluated (our_version >= min_major = satisfied)
3. Add fallback detection: when no Dockerfile or engines.node is found,
check .nvmrc and .node-version files for Node version hints
4. Add subdirectory search: Dockerfile and package.json are now found
via GitHub API tree search, not just in repo root
5. Use GitHub API to detect default branch instead of guessing
main/master/dev with multiple HEAD requests
* fix typos in node_version
* runs on vps
* fix(error-handler): prevent silent() from re-enabling error handling during recovery
Root cause: silent() (core.func) unconditionally calls set -Eeuo pipefail
and trap 'error_handler' ERR after every command. When build_container()
intentionally disables error handling for its recovery section, any
intermediate call through silent()/ re-enables it. This causes the
grep/sed pipeline for missing_cmd extraction to trigger error_handler
(grep returns exit code 1 on no match + pipefail = fatal).
Fixes:
1. silent(): Save errexit state before disabling, only restore if it was
active. Callers that intentionally disabled error handling (e.g.
build_container recovery) are no longer silently re-enabled.
2. build.func: Add || true to missing_cmd grep pipeline as defense-in-depth
against pipeline failure propagation.
3. build.func: Add explicit set +Eeuo pipefail / trap - ERR after
post_update_to_api() call, before error classification grep/sed section.
4. build.func: Remove stale global combined_log variable from variables()
that used a different path format (/tmp/install-SESSION-combined.log)
than the actual local variable (/tmp/NSAPP-CTID-SESSION.log). The global
was never written to and caused confusion when error_handler displayed it.
* Update build.func
* chore(install): add Github source links to all setup_nodejs scripts
52 install scripts had a project website in '# Source:' but no GitHub
link. Merged the GitHub repo URL into the Source header as:
# Source: https://website.com/ | Github: https://github.com/OWNER/REPO
Repos sourced from fetch_and_deploy_gh_release calls, get_latest_github_release
calls, or known project repos for npm/pip installed apps.
Two scripts (fumadocs, pve-scripts-local) had no Source line at all —
added one. Shinobi skipped (GitLab-only, no GitHub repo).
* chore(install): add Github source links to all fetch_and_deploy scripts
77 additional install scripts had fetch_and_deploy_gh_release calls but
no GitHub link in the Source header. Merged the primary app repo into
the Source header as:
# Source: https://website.com/ | Github: https://github.com/OWNER/REPO
Where multiple fetch_and_deploy calls existed (app + dependency), the
primary app repo was selected:
- ersatztv: ErsatzTV/ErsatzTV (not ffmpeg)
- firefly: firefly-iii/firefly-iii (not data-importer)
- komga: gotson/komga (not kepubify dep)
- sabnzbd: sabnzbd/sabnzbd (not par2cmdline-turbo dep)
- signoz: SigNoz/signoz (not otel-collector)
- tunarr: chrisbenincasa/tunarr (not ffmpeg dep)
Also fixed cosmos-install.sh double https:// in Source URL.
Skipped: autocaliweb (source already on codeberg, GitHub repos are deps only)
* revert: restore misc/build.func and misc/core.func to main state
These error-handler fixes belong to fix/error-handler-recovery, not to
this sources-only branch.
* chore(ct,tools): sync Source headers with install/ and add Github links to addon scripts
Scans all install scripts using setup_nodejs and compares our
NODE_VERSION with upstream Dockerfile and package.json values.
Features:
- Detects FROM node:XX, nodesource/setup_XX, FROM alpine:X.Y
- Resolves Alpine package registry for nodejs version when
upstream uses alpine base images
- Caches Alpine version lookups to minimize requests
- Creates individual GitHub issues per script with investigation
checklist when drift is detected
- Rate-limited to avoid GitHub API throttling
- Runs weekly on Monday at 06:00 UTC + manual dispatch
Adds a reusable function to fetch the latest tag from a GitHub repo.
Useful for projects that only use tags, not full releases (e.g.
mongodb/mongo-tools).
Features:
- Optional prefix filter (e.g. '100.' or 'v')
- Optional prefix stripping for clean version output
- Skips pre-release tags (rc, alpha, beta, dev, test)
- Sorts by version (sort -V) to find the latest
- Respects GITHUB_TOKEN for rate limiting
Root cause: post_update_to_api set POST_UPDATE_DONE=true even after
all 3 retry attempts failed (curl timeout, API error). This prevented
the EXIT trap (api_exit_script) from retrying with fresh attempts.
Changes:
- Only set POST_UPDATE_DONE=true on actual HTTP 2xx success
- If all 3 attempts fail, EXIT trap gets 3 more fresh attempts
- Increase timeout from 5s to 10s for final status updates (STATUS_TIMEOUT)
Progress pings keep 5s (TELEMETRY_TIMEOUT) since they're lightweight
- post_update_to_api_extended: add proper retry logic + HTTP code check
(was fire-and-forget with no retry)
Validation status was persisting through container start, network check,
and OS customization. Now transitions to 'configuring' immediately after
create_lxc_container returns. Validation only covers storage/template/cluster
checks as intended.
Pipe lxc-attach output through tee into /tmp/.install-capture-${SESSION_ID}.log and use PIPESTATUS[0] to get the real lxc-attach exit code. Prefer a pulled container-side INSTALL_LOG when it exists and is >100 bytes; otherwise fall back to the host-captured terminal log (stripping ANSI codes) and append it to the combined log so get_full_log() can find it. Apply the same capture behavior to the retry path and remove temporary capture files on completion. This makes install output reliable when container-side logging is missing (DNS errors, early crashes, or missing silent() usage).
After 'export INSTALL_LOG' in build.func, get_active_logfile() returned
the container's INSTALL_LOG path for all host-side logging, causing
msg_info/msg_ok/msg_error on the host to write to /root/.install-SESSION.log
(the host file, not the container's) instead of BUILD_LOG. This made
BUILD_LOG incomplete and get_full_log() unable to send full traces.
Fix: Add _HOST_LOGFILE (not exported, invisible to container) so the host
always logs to BUILD_LOG. Container still uses INSTALL_LOG as before.
* Enhance telemetry, signal handling, and logs
Improve failure telemetry and signal handling across the installer: add get_full_log() to collect/strip/truncate install logs and include them in API payloads with a truncated retry; add CONTAINER_INSTALLING flag around lxc-attach and stop containers on abort to avoid orphaned "installing/configuring" records; introduce _send_abort_telemetry() (curl fallback for container context) and _stop_container_if_installing() helpers; centralize and simplify EXIT/ERR/INT/TERM/HUP traps and handlers (including a new on_hangup handler) and update VM scripts to report numeric exit codes. Also ensure best-effort log collection is performed and tweak error categorization for certain signals.
* Include full log in error telemetry
Use get_full_log (up to 120KB) to populate the error telemetry field so the API receives the full installation trace; fall back to get_error_text (last ~20 lines) if the full log is empty. Removed collection and inclusion of a separate install_log field from the JSON payloads and simplified the retry payloads/comments accordingly. The change ensures error reports contain the complete trace while avoiding duplicate large log fields and keeps graceful failure handling (get_full_log || true).
* Anonymize IP addresses in get_full_log
Mask IPv4 addresses in logs when collecting full log output: added a sed step that replaces the last two octets with "x.x" to avoid exposing full IPs (GDPR). Also updated the comment to reflect anonymization; existing steps that strip carriage returns and ANSI escape sequences remain in place before truncating with head -c.
Prevent host-side error_handler from being triggered during in-container install/recovery by delaying re-enabling set -Eeuo pipefail and the ERR trap in misc/build.func until after install/recovery completes; add explanatory comments. Update misc/error_handler.func to fall back to BUILD_LOG if container-internal log path is unavailable, show the last 20 log lines when present, refine container vs host detection (check INSTALL_LOG file and /root), copy INSTALL_LOG into /root and write a .failed flag with the exit code for host-side detection, and ensure full-log output and container removal prompt are shown appropriately in host context. Tweak misc/core.func silent() output to include a "Full log" path and adjust formatting.
Update misc/tools.func to include Signed-By: /usr/share/keyrings/debian-archive-keyring.gpg on multiple Debian repository blocks (bullseye, bookworm, trixie/sid and their -security entries) for non-free and non-free-firmware components. This ensures APT uses the specified keyring to verify repository metadata and improves reproducible/secure apt configuration.
* Update frigate.sh
* Add Frigate metadata and overhaul installer
Add frontend metadata for Frigate (frontend/public/json/frigate.json), remove the old .bak metadata file, and perform a major refactor of install/frigate-install.sh. The installer now targets Debian 12 (Bookworm), converts APT sources to deb822, installs and builds required dependencies (Python wheels, libusb, OpenVINO, Tempio, Nginx, sqlite extensions), configures hardware acceleration and GPU access, fetches and deploys Frigate and go2rtc releases, and prepares inference/audio models. Systemd service units were improved (dependencies, env file usage, safer log file handling, create_directories service) and services are enabled/started with cleanup steps added. Also updated copyright/authorship and various runtime environment exports and default Frigate config changes (ffmpeg hwaccel, detector selection, disabled auth/detect in default config).
* Update frigate.json
* frigate: update metadata and installer
Update frontend metadata (config path, interface port, and expanded description) and modernize the install script: switch apt-get to apt, streamline dependency list (remove wget/jq/unzip), replace inline hardware-acceleration/GPU group tweaks with setup_hwaccel, pin Frigate release to v0.16.4 for reproducible installs, and fetch/libusb build now uses fetch_and_deploy_gh_release with adjusted paths. Also clean up removed temporary files.
* add std
- check_for_gh_release: add Authorization header when GITHUB_TOKEN is set,
detect HTTP 403 and show actionable rate-limit hint
- fetch_and_deploy_gh_release: improve retry loop with specific 403 handling,
exponential backoff, and token export hint on rate-limit failure
npm run db:push ran drizzle migrate (CREATE TABLE) on an empty directory
before the config backup was restored, creating a fresh DB that got
immediately overwritten by the old backup. The old DB was missing new
columns like resources.postAuthPath (added in 1.15.4).
Replace with drizzle-kit push after config restore, which introspects the
existing DB and applies schema diffs (ALTER TABLE) instead.
Ref: #12068
Recyclarr v8.0 moved local includes from configs/ to a dedicated
includes/ directory. The install script now creates both dirs
upfront. The update script migrates existing include subdirs
from configs/ to includes/ after the binary update and also
handles detection for both old (recyclarr.yml) and new
(configs/) config layouts.
See https://recyclarr.dev/guide/upgrade-guide/v8.0
Ref #12109
Planka v2.0.1 moved all uploads to a unified data directory:
- public/{favicons,user-avatars,background-images} -> data/protected/
- private/attachments -> data/private/
The update script was still restoring files to the old locations,
which made attachments, avatars and backgrounds disappear after
updating from rc.4 to 2.0.1.
Changes:
- update_script: backup handles both old and new layouts
- update_script: restore now uses new v2 paths
- install script: create v2 data directories on fresh install
Ref #12066
The JVM heap options were appended with >> which left the old
defaults in place. Use sed to replace them instead.
Also add the missing systemctl daemon-reload that ES requires
after package installation before the service can be started.
Closes#12111
* fix(huntarr): add build-essential for native pip dependencies
Some Python packages in requirements.txt need a C++ compiler for
native extensions. Without build-essential the uv pip install
step fails with 'command c++ not found'.
Closes#12117
* Change apt-get to apt for installing dependencies
* Ensure git dependency before updating Huntarr
Added dependency check for git before fetching and deploying the GitHub release.
* drunk
* Fix: Add disk space checking for OPNsense VM FreeBSD image decompression
- Add check_disk_space() function to verify available storage
- Check for 20GB before download and 15GB before decompression
- Provide clear error messages showing available vs required space
- Add proper error handling for unxz decompression failures
- Clean up compressed .xz file after decompression to save space
- Add progress messages for download and decompression steps
Fixes issue where script fails at line 611 with 'No space left on device'
when /tmp directory lacks sufficient space for ~10-15GB decompressed image.
* Increase OPNsense VM disk size from 10GB to 20GB
- Provides more space for system updates, logs, and package installations
- 20GB is a more appropriate size for OPNsense production use
Introduce post_progress_to_api() in misc/api.func — a non-blocking, fire-and-forget curl ping (gated by DIAGNOSTICS and RANDOM_UUID) that updates telemetry status to "configuring". Wire this progress ping into multiple scripts (alpine-install.func, install.func, build.func, core.func) at key milestones (container start, network ready, customization, creation, cleanup) and replace/deduplicate some earlier post_to_api calls. Also update error_handler.func to always report failures immediately via post_update_to_api to ensure failures are captured even before/after container lifecycle.
Prevent hangs when pulling logs from containers by wrapping pct pull calls with timeout (8s) and running ensure_log_on_host under timeout (10s). Always send telemetry (post_update_to_api) before attempting best-effort log collection so status is reported even if log retrieval blocks. Update EXIT/ERR/SIGHUP/SIGINT/SIGTERM traps and consolidate error/interrupt handlers to use the new timeouted log collection. Changes in misc/build.func and misc/error_handler.func.
* fix: send telemetry BEFORE log collection in signal handlers
- Swap ensure_log_on_host/post_update_to_api order in on_interrupt, on_terminate, api_exit_script, and inline SIGHUP/SIGINT/SIGTERM traps
- For signal exits (>128): send telemetry immediately, then best-effort log collection
- Add 2>/dev/null || true to all I/O in signal handlers to prevent SIGPIPE
- Fix on_exit: exit_code=0 now reports 'done' instead of 'failed 1'
- Root cause: pct pull hangs on dying containers blocked telemetry updates, leaving 595+ records stuck in 'installing' daily
* feat: add execution_id to all telemetry payloads
- Generate EXECUTION_ID from RANDOM_UUID in variables()
- Export EXECUTION_ID to container environment
- Add execution_id field to all 8 API payloads in api.func
- Add execution_id to post_progress_to_api in install.func and alpine-install.func
- Fallback to RANDOM_UUID when EXECUTION_ID not set (backward compat)
* fix: correct telemetry type values for PVE and addon scripts
- PVE scripts (tools/pve/*): change type 'tool' -> 'pve'
- Addon scripts (tools/addon/*): fix 4 scripts that wrongly used 'tool' -> 'addon'
(netdata, add-tailscale-lxc, add-netbird-lxc, all-templates)
- api.func: post_tool_to_api sends type='pve', default fallback 'pve'
- Aligns with PocketBase categories: lxc, vm, pve, addon
* fix: persist diagnostics opt-in inside containers for addon telemetry
- install.func + alpine-install.func: create /usr/local/community-scripts/diagnostics
inside the container when DIAGNOSTICS=yes (from build.func export)
- Enables addon scripts running later inside containers to find the opt-in
- Update init_tool_telemetry default type from 'tool' to 'pve'
* refactor: clean up diagnostics/telemetry opt-in system
- diagnostics_check(): deduplicate heredoc (was 2x 22 lines), improve whiptail
text with clear what/what-not collected, add telemetry + privacy links
- diagnostics_menu(): better UX with current status, clear enable/disable
buttons, note about existing containers
- variables(): change DIAGNOSTICS default from 'yes' to 'no' (safe: no
telemetry before user consents via diagnostics_check)
- install.func + alpine-install.func: persist BOTH yes AND no in container
so opt-out is explicit (not just missing file = no)
- Fix typo 'menue' -> 'menu' in config file comments
* fix: no pre-selection in telemetry dialog, link to telemetry-service README
- Add --defaultno so 'No, opt out' is focused by default (user must Tab to Yes)
- Change privacy link from discussions/1836 to telemetry-service#privacy--compliance
* fix: use radiolist for telemetry dialog (no pre-selection)
- Replace --yesno with --radiolist: user must actively SPACE-select an option
- Both options start as OFF (no pre-selection)
- Cancel/Exit defaults to 'no' (opt-out)
* simplify: inline telemetry dialog text like other whiptail dialogs
* improve: telemetry dialog with more detail, link to PRIVACY.md
- Add what we collect / don't collect sections back to dialog
- Link to telemetry-service/docs/PRIVACY.md instead of README anchor
- Update config file comment with same link
* core: add progress; fix exit status
Introduce post_progress_to_api() in alpine-install.func and install.func to send a lightweight, fire-and-forget telemetry ping (HTTP POST) that updates an existing telemetry record to "configuring" when DIAGNOSTICS=yes and RANDOM_UUID is set. The function is non-blocking (curl -m 5, errors ignored) and is invoked during container setup and after OS updates to signal active progress. Also adjust api_exit_script() in build.func to report success (post_update_to_api "done" "0") for cases where the script exited normally but a completion status wasn't posted, instead of reporting failure.
* Safer tools.func load and improved error handling
Replace process-substitution sourcing of tools.func with an explicit curl -> variable -> source via /dev/stdin, adding failure messages and a check that expected functions (e.g. fetch_and_deploy_gh_release) are present (misc/alpine-install.func, misc/install.func). Add categorize_error mapping for exit code 10 -> "config" (misc/api.func). Tweak build.func: minor pipeline formatting and change the ERR trap to capture the actual exit code and only call ensure_log_on_host/post_update on non-zero exits, preventing erroneous failure reporting.
* tools: add data init and auto-reporting to tools and pve section
Introduce telemetry helpers in misc/api.func: _telemetry_report_exit (reports success/failure via post_tool_to_api/post_addon_to_api) and init_tool_telemetry (reads DIAGNOSTICS, starts install timer and installs an EXIT trap to auto-report). Integrate telemetry into many tools/addon and tools/pve scripts by sourcing the remote api.func and calling init_tool_telemetry (guarded with declare -f). Also apply a minor arithmetic formatting tweak in misc/build.func for RECOVERY_ATTEMPT.
* core: add progress; fix exit status
Introduce post_progress_to_api() in alpine-install.func and install.func to send a lightweight, fire-and-forget telemetry ping (HTTP POST) that updates an existing telemetry record to "configuring" when DIAGNOSTICS=yes and RANDOM_UUID is set. The function is non-blocking (curl -m 5, errors ignored) and is invoked during container setup and after OS updates to signal active progress. Also adjust api_exit_script() in build.func to report success (post_update_to_api "done" "0") for cases where the script exited normally but a completion status wasn't posted, instead of reporting failure.
* Safer tools.func load and improved error handling
Replace process-substitution sourcing of tools.func with an explicit curl -> variable -> source via /dev/stdin, adding failure messages and a check that expected functions (e.g. fetch_and_deploy_gh_release) are present (misc/alpine-install.func, misc/install.func). Add categorize_error mapping for exit code 10 -> "config" (misc/api.func). Tweak build.func: minor pipeline formatting and change the ERR trap to capture the actual exit code and only call ensure_log_on_host/post_update on non-zero exits, preventing erroneous failure reporting.
Introduce create_service() to generate the immich-proxy systemd unit and run systemctl daemon-reload. Replace duplicated heredoc service blocks in install with a call to create_service, and invoke create_service during update before starting the service. Adjust unit WorkingDirectory to ${INSTALL_PATH}/app and ExecStart to run dist/index.js.
* Ensure API update is sent on script exit
Add exit-time telemetry handling across scripts to avoid orphaned "installing" records. Introduce local exit_code capture in api_exit_script and cleanup handlers and, when POST_TO_API_DONE is true but POST_UPDATE_DONE is not, post a final status (marking failures on non-zero exit codes, or marking done/failed in VM cleanups based on exit code). Changes touch misc/build.func, misc/vm-core.func and various vm/*-vm.sh cleanup functions to reliably send post_update_to_api on normal or early exits.
* Update api.func
* fix(telemetry): add missing exit codes to explain_exit_code()
- Add curl error codes: 4, 5, 8, 23, 25, 30, 56, 78
- Add code 10: Docker/privileged mode required (used in ~15 scripts)
- Add code 75: Temporary failure (retry later)
- Add BSD sysexits.h codes: 64-77
- Sync error_handler.func fallback with canonical api.func
Replace the final '| head -1' in both install and ct scripts with 'awk 'NR==1'' to pick the first matching Teamspeak release line. In the ct script the previous temporary toggling of pipefail was also removed, simplifying the command. This improves compatibility and reduces reliance on an extra utility in minimal environments.
- Remove legacy error_handler(), on_exit(), on_interrupt(), on_terminate() and set/trap definitions from alpine-install.func (already provided by error_handler.func which is sourced on line 10)
- The local error_handler() expected positional args as required, but catch_errors() sets trap as 'error_handler' (without args), causing unbound variable error with set -u (nounset)
- error_handler.func uses default values which is set -u safe
- Also align legacy trap in install.func network_check() to standard format
Fixes#11929
Update ct/jellyseerr.sh and ct/overseerr.sh to switch the container update handler to the Seerr script. The here-doc now uses a single-quoted EOF to avoid shell expansion and includes an explicit shebang for the generated /usr/bin/update. Instead of auto-executing the new update script, the code now informs the user to run 'update' again and exits (overseerr exits with 0). Also includes minor whitespace cleanup (removed trailing spaces on cd lines). This prevents unexpected immediate execution and ensures the generated script runs with the intended shell.
- Add 'exit 130' after SIGINT trap handler
- Add 'exit 143' after SIGTERM trap handler
- Fixes issue where interrupted scripts stayed as 'Installing' in telemetry
- Previously the traps only sent the update but didn't terminate the script
- Add || true to dmidecode pipelines to prevent script abort when
'Configured Memory Speed: Unknown' is returned (no numeric match)
- Fixes#11913 edge case for nested ProxmoxVE VMs
Update hardcoded fallback REPO_SOURCE from ProxmoxVED to ProxmoxVE and clarify comment about production vs dev repository naming. Add fail-safe '|| true' to several detection pipelines (lspci for GPU and grep commands reading /proc/cpuinfo for CPU vendor/model) to avoid non-zero exits or empty outputs causing function failures and improve robustness.
- Add || true to dmidecode pipelines to prevent error when speed is 'Unknown'
- Validate RAM_SPEED is a valid integer, fallback to 0
- Send ram_speed as integer (not string) in all JSON payloads for PocketBase
Fixes: dmidecode in nested VMs returns 'Configured Memory Speed: Unknown'
which causes grep to fail and triggers catch_errors handler.
Enhance json_escape to remove ANSI escape sequences (color codes) and other non-printable control characters before escaping backslashes, quotes, newlines, tabs and carriage returns. Also update get_error_text to strip ANSI sequences from tailed logfile output. These changes ensure safe JSON embedding of strings and prevent control characters / terminal color codes from leaking into logs or JSON payloads.
- Query /cluster/resources via pvesh to check all VMs/CTs on ALL nodes
- Check /etc/pve/nodes/*/qemu-server and /etc/pve/nodes/*/lxc dirs
- Handles pmxcfs sync delays that caused sporadic ID conflicts
- Remove duplicate validate_container_id/get_valid_container_id definitions
- Add max_attempts safeguard to prevent infinite loops
* fix(telemetry): improve error reporting with structured error strings and better categorization
- Add build_error_string() that creates structured format:
'exit_code=N | description\n---\n<last 20 log lines>'
- Fix categorize_error() to map ALL known exit codes:
- Added: shell(1,2), proxmox(200-231), service(150-154),
database(170-193), runtime(243-249), signal(139,141,143)
- Split timeout from network (28 was in both)
- Added DPKG(255) to dependency category
- Update all API functions to use build_error_string():
post_update_to_api, post_update_to_api_extended,
post_tool_to_api, post_addon_to_api
- Add ensure_log_on_host() calls to on_exit, on_interrupt,
on_terminate handlers to prevent race condition where
telemetry reports before container log is pulled to host
* fix(ui): improve error output formatting and remove redundant log paths
- error_handler: Use msg_info/msg_ok/msg_warn for container cleanup
instead of raw echo with manual ANSI codes
- error_handler: Add ❓ icon before 'Remove broken container?' prompt
- error_handler: Indent log output with TAB for visual consistency
- build.func: Use msg_custom for installation log path display
- build.func: Use msg_info → msg_ok for container removal flow
- build.func: Use msg_warn for 'kept for debugging' message
- core.func/vm-core.func: Remove redundant container-internal log
path display (📋 View full log) since combined log on host is
the canonical location shown after failure
Apply json_escape to GPU_MODEL and CPU_MODEL before assigning to gpu_model and cpu_model to ensure values are safe for inclusion in API JSON payloads. Updated in post_to_api, post_to_api_vm, and post_update_to_api; variable declarations were adjusted to call json_escape on the existing environment values (fallbacks unchanged). This prevents raw model strings from breaking the API payload.
Enhance post_update_to_api to support a "force" mode and robust retry logic: add a 3rd-arg bypass to duplicate suppression, capture a short error summary, and perform up to three POST attempts (full payload, shortened error payload, minimal payload) with HTTP code checks and small backoffs. Mark POST_UPDATE_DONE on success (or after three attempts) to avoid infinite retries. Also invoke post_update_to_api with the "force" flag from cleanup paths in build.func and error_handler.func so a final status update is attempted after cleanup.
Copy the container install log to a host path before reporting a failure to the telemetry API so get_error_text() can read it. Introduce host_install_log and point INSTALL_LOG to the host copy when pulled via pct, move post_update_to_api after the log copy, and update the displayed installation-log path.
Update scripts to use Python 3.12 for uv tool setup and Open-WebUI installs/upgrades. Add a numba constraint (--constraint <(echo "numba>=0.60")) to uv tool install/upgrade commands to ensure compatibility. Changes applied to ct/openwebui.sh and install/openwebui-install.sh for both fresh installs and update paths.
Replace npm script calls to db:sqlite:generate and db:sqlite:push with db:generate and db:push in ct/pangolin.sh and install/pangolin-install.sh. This makes the build/install steps use the generic DB task names for consistency across update and install workflows.
* fix(archlinux-vm): fix LVM/LVM-thin storage and improve error reporting
- Add catch-all (*) case for storage types (LVM, LVM-thin, zfspool)
Previously only nfs/dir/cifs and btrfs were handled, leaving
DISK_EXT, DISK_REF, and DISK_IMPORT unset on LVM/LVM-thin storage
- Fix error_handler to send numeric exit_code to API instead of
bash command text (which caused 'Unknown error' in telemetry)
- Replace fragile pvesm alloc for EFI disk with Proxmox-managed
:0,efitype=4m (consistent with docker-vm.sh)
- Modernize disk import: auto-detect qm disk import vs qm importdisk,
parse output to get correct disk reference instead of guessing names
- Use --format flag (double dash) consistent with modern Proxmox API
- Remove unused FORMAT variable (EFI type now always set correctly)
- Remove fragile eval-based disk name construction
* fix(vm): fix LVM/LVM-thin storage and error reporting for all VM scripts
- Add catch-all (*) case to storage type detection in all VM scripts
that were missing it (debian-vm, debian-13-vm, ubuntu2204/2404/2504,
nextcloud-vm, owncloud-vm, opnsense-vm, pimox-haos-vm)
- Add catch-all to mikrotik-routeros (had zfspool but not lvm/lvmthin)
- Fix error_handler in ALL 14 VM scripts to send numeric exit_code
to post_update_to_api instead of bash command text, which caused
'Unknown error' in telemetry because the API expects a number
The catch_errors() function in CT scripts overrides the API telemetry
traps set by build.func. This caused on_exit, on_interrupt, and
on_terminate to never call post_update_to_api, leaving telemetry
records permanently stuck on 'installing'.
Changes:
- on_exit: Report orphaned 'installing' records on ANY exit where
post_to_api was called but post_update_to_api was not
- on_interrupt: Call post_update_to_api('failed', '130') before exit
- on_terminate: Call post_update_to_api('failed', '143') before exit
All calls are guarded by POST_UPDATE_DONE flag to prevent duplicates.
EMQX 6.1+ preallocates significant disk space for the MQ feature,
causing high CPU/disk usage on small containers (emqx/emqx#16649).
- Increase default disk from 4GB to 6GB
- Add read -rp prompt during install to optionally disable MQ feature
via mq.enable=false in emqx.conf (reduces disk/CPU overhead)
- Setting is in install script (not CT script) per reviewer feedback
Co-authored-by: sim-san <sim-san@users.noreply.github.com>
The UniFi GPG key at dl.ui.com/unifi/unifi-repo.gpg is already in binary
format. setup_deb822_repo unconditionally ran gpg --dearmor which expects
ASCII-armored input, corrupting binary keys and causing apt to fail with
'Unable to locate package unifi'.
setup_deb822_repo now downloads the key to a temp file first and uses
the file command to detect whether it is already a binary PGP/GPG key.
Binary keys are copied directly; armored keys are dearmored as before.
This also reverts unifi-install.sh back to using setup_deb822_repo for
consistency with all other install scripts.
* fix(tailscale-addon): fix DNS check and keyrings directory issues
- Source /etc/os-release instead of grep to handle quoted values properly
- Use VERSION_CODENAME variable instead of VER for correct URL
- Add fallback DNS resolution methods (host, nslookup, getent) when dig is missing
- Create /usr/share/keyrings directory if it doesn't exist
- Skip DNS check gracefully when no DNS tools are available
Fixes installation failures with 'dig: command not found' and
'No such file or directory' for keyrings path
* Update tools/addon/add-tailscale-lxc.sh
Co-authored-by: Chris <punk.sand7393@fastmail.com>
* Update tools/addon/add-tailscale-lxc.sh
Co-authored-by: Chris <punk.sand7393@fastmail.com>
---------
Co-authored-by: Chris <punk.sand7393@fastmail.com>
Add POST_TO_API_DONE idempotency guard to post_to_api() to prevent
the same telemetry record from being submitted twice with the same
RANDOM_UUID. This mirrors the existing POST_UPDATE_DONE pattern in
post_update_to_api().
post_to_api() is called twice in build.func:
- After storage validation (inside CONTAINER_STORAGE check)
- After create_lxc_container() completes
When both execute, the second call fails with a random_id uniqueness
violation on PocketBase, generating server-side errors.
* fix(debian-13-vm): disable systemd-firstboot and add autologin via virt-customize
Newer Debian 13 (Trixie) nocloud images (Jan 29+) have systemd-firstboot
enabled, which blocks the boot process by prompting for timezone and root
password on the serial console. The noVNC console stalls after auditd checks.
Fix by using virt-customize (like docker-vm.sh) to:
- Disable systemd-firstboot.service and mask it
- Pre-seed timezone to Etc/UTC
- Configure autologin on ttyS0 and tty1 for nocloud images
- Set hostname and clear machine-id for unique IDs
- Install libguestfs-tools if not present
Fixes#11807
* feat(debian-13-vm): add cloud-init selection dialog for default and advanced settings
Add select_cloud_init() function consistent with docker-vm.sh that prompts
the user to choose between Cloud-Init (genericcloud image) and nocloud
(with auto-login) in both default and advanced settings.
Previously, default settings hardcoded CLOUD_INIT=no without asking.
* Remove Go API and extend misc/api.func
Delete the Go-based API (api/main.go, api/go.mod, api/go.sum, api/.env.example) and significantly enhance misc/api.func. The shell telemetry file now includes telemetry configuration, repo source detection, GPU/CPU/RAM detection, expanded explain_exit_code mappings, and refactored post_to_api/post_to_api_vm to send non-blocking telemetry to telemetry.community-scripts.org while respecting DIAGNOSTICS/DEV_MODE and adding richer metadata (cpu/gpu/ram/repo_source). Also updates header/author info and improves privacy/robustness and error handling.
* Start install timer and refine error reporting
Call start_install_timer during build startup and overhaul exit/error reporting.
Changes:
- Invoke start_install_timer early in misc/build.func to track install duration.
- Update api_exit_script comments to reference PocketBase/api.func and adjust ERR/SIGINT/SIGTERM traps to post numeric exit codes (use $? / 130 / 143) instead of command strings.
- Replace the previous explain_exit_code implementation with a conditional fallback: only define explain_exit_code if not already provided (api.func is the canonical source). Expanded and reorganized exit code mappings (curl, timeout, systemd, Node/Python/Postgres/MySQL/MongoDB, Proxmox, etc.).
- In error_handler: stop echoing the container log path (host shows combined log), and post a "failed" update to the API with the exit code before offering container cleanup.
Rationale: these changes make telemetry more consistent and robust (numeric codes), provide a safe fallback for exit descriptions when api.func isn't loaded, and ensure failures are reported to the API prior to any automatic cleanup.
* Report install start/failure to telemetry API
Add telemetry hooks in misc/build.func: call post_to_api at installation start to capture early or immediately-failing installs, and call post_update_to_api with status "failed" and the install exit code when a container installation fails. This improves visibility into install failures for monitoring/telemetry.
Kasm changed their release URL format starting with v1.18.1:
- Old format: kasm_release_1.18.0.09f70a.tar.gz (with hash)
- New format: kasm_release_1.18.1.tar.gz (without hash)
The script now tries both detection methods:
1. First tries to find URL with hash suffix (old format)
2. Falls back to detecting version from service_images URLs and
constructing the new URL format
This fixes the update detection for Kasm v1.18.1 and future versions.
Fixes#11785
CanbiZ (MickLesk)