Two hardcoded event-triggered entries appear at the top of the scheduler
(array_started.sh / array_stopping.sh). They show an ⚡ Array Start /
⚡ Array Stop badge instead of a cron field and are enabled/disabled via
the normal toggle.
Static event scripts fire them via Unraid's event system:
- event/disks_mounted/array_start_jobs → runs in background (non-blocking)
- event/disks_unmounting/array_stop_jobs → runs foreground (blocks until done)
vv_cron_rebuild() skips @array_* entries so they never land in the cron file.
Scripts executed on each event are managed in master.conf via
ARRAY_START_SCRIPTS and ARRAY_STOP_SCRIPTS arrays.
Auto Scroll now pauses automatically when user scrolls up and resumes
when they reach the bottom — no separate Scroll Lock needed.
Lidarr discovery no longer hard-skips single-seed candidates; they
score low on breadth but can still reach threshold on merit.
Before update_cron was wired in, the plugin wrote directly to
/etc/cron.d/varaverk. That file persisted after the migration and
caused duplicate job firing (two watchdog instances at :00/:15/:30/:45).
The second instance hit the lock, exited with warn, and overwrote the
stat file — killing the running indicator while the first was still live.
vv_cron_rebuild() now unlinks the legacy file after every update_cron
call. Since rebuild_cron (boot event) calls vv_cron_rebuild(), the
cleanup runs on every reboot and every schedule save.
run_job.sh was overwriting the log on every cron fire, making it
impossible to see run history. Switch to append mode, print a
── YYYY-MM-DD HH:MM:SS ── separator before each run, and trim to
the last 1000 lines after each run to keep logs bounded.
Cron runs with PATH=/usr/bin:/bin, which omits /usr/local/sbin where
tailscale (and other Unraid tools) live. resolve_remote_ip silently got
empty results from both tailscale calls, failing all 3 retry attempts
every scheduled run. Export the full path at source time so all scripts
find tailscale regardless of how they're invoked.
Scripts calling acquire_lock followed by acquire_rsync_lock (rsync.sh)
or a custom EXIT trap (resource_watchdog.sh) would overwrite the lock
release trap, leaving the first lock file behind on exit.
Replace per-acquire trap with a _LOCK_FILES registry. _register_lock()
appends each lock file to the array and sets a single _release_all_locks
trap. All acquired locks release together on exit regardless of how many
traps are subsequently set or overwritten.
resource_watchdog.sh: chain _release_all_locks into both its custom trap
and the state-persisted disarm path so the lock releases in all exit paths.
The first cron run of critical_sync_maintenance failed because Tailscale
had a transient moment at startup. resolve_remote_ip() now retries 3 times
with a 5s delay before giving up, making it resilient to brief daemon
hiccups without masking real peer-offline conditions.
Add api/status.php — reads all scheduled jobs' stat files and returns a
{id: status} map. The scheduler page polls it every 3s and lights up the
dot on any job whose stat shows status=running, clearing it when done.
Replace single vvRunningId with vvRunningSet (Set) so multiple jobs can
show as running simultaneously. vvSetDot/vvClearDot handle per-job dots
and the header dot for the active job. Remove stale-count detection from
vvFetchRight — status poll is now the single source of truth.
Create varaverk.plg on flash (/boot/config/plugins/varaverk/varaverk.plg).
On boot, rebuild_cron copies it to /var/log/plugins/ so update_cron natively
picks up varaverk.cron and merges it into /etc/cron.d/root.
With the .plg in place the dev-mode direct-inject fallback in vv_cron_rebuild()
is no longer needed and has been removed.
dcron uses the filename in /etc/cron.d/ as the username, so a file named
'varaverk' ran jobs as a non-existent user and was silently ignored.
Switch to writing /boot/config/plugins/varaverk/varaverk.cron and calling
update_cron (which merges all plugin .cron files into /etc/cron.d/root).
In dev mode (no varaverk.plg registered), fall back to injecting our entries
directly into the live root crontab via crontab -c /etc/cron.d -.
run_job.sh: new wrapper called by scheduler cron for every job.
Writes /var/log/varaverk/<id>.json (status/start/end/exit/pid)
and /var/log/varaverk/<id>.log (latest run output, overwritten).
Maps exit 0→ok, 1→warn, 2+→error so watchdog "took action" runs
show warn not error.
scheduler: cron now calls bash run_job.sh instead of inline bash -c.
Added vv_job_stat_path(). LOG_DIR moved to config.php (shared).
monitor: vv_scripts_status() now reads /var/log/varaverk/*.json and
*/*.json — no User Scripts tmpScripts dependency at all.
Stale running detection via /proc/<pid> check.
Returns error_count in addition to running/ok/warn counts.
pages/monitor.php: added error pill (red ✗), error icon/color in
list, duration shown next to age for each script entry.
log() is gated on ENABLE_LOGGING — silent on normal runs. Healthy
confirmations (All healthy, Network healthy, Storage healthy, WebGUI
healthy) were invisible, making banners appear with nothing after them.
Switched all four to plain echo so they show every cycle.
All watchdog scripts now open with a consistent echo banner:
━━━ <icon> <Name> Watchdog — <timestamp> ━━━
Previously network and storage used a quiet log() line (no separator),
and stability_watchdog used warn() which made every run show ⚠️.
Also: network_watchdog now logs "Network healthy ✅" on clean exit,
and stability_watchdog's reboot-trigger block now says "Stability"
instead of the old "System Watchdog" label.
network_watchdog: Tailscale JSON uses "BackendState": "Running" (with space)
but grep pattern matched "BackendState":"Running" (no space) — always failed.
Switched to grep -qE with \s* to handle both forms.
system_watchdog: add echo banner at run start so its output is visually
separated from docker_watchdog in orchestrator logs. Previously appeared
as a continuation of the docker section since docker_watchdog has a banner
but system_watchdog did not.
Layout:
- Switch from three flex rows to single CSS grid (repeat(8,1fr)) — eliminates
gap-count discrepancy between rows; span 1/2/4/8 cards are now pixel-identical
across all rows regardless of how many cards share the row
New cards:
- Scripts card: last-run status (ok/warn/running/unknown), clickable filter pills,
7-row scroll, flex:1 matching Power/Parity
- Streams card: live per-second progress counters, color-coded play buttons
(blue=live, orange=transcoding, green=direct, yellow=paused), server badges
with counts, left-fill column layout (3 per col)
- Transcode card: NVMe indicator, active session count + scrollable list,
session rows with server icon + title + type + method
Dynamic thresholds:
- vv_disk_thresholds() reads /boot/config/plugins/dynamix/dynamix.cfg
- Disk utilization bars: green/orange/red based on warning/critical settings
- Temperature colors: hot/max for HDD, hotssd/maxssd for SSD/NVMe
- Thresholds update every poll cycle; changing Unraid settings takes effect automatically
Other:
- SSD path detection uses findmnt FSTYPE to exclude tmpfs/ramfs (fixes RAM
being reported as SSD transcode target)
- varaverk.css h3: add white-space:normal + overflow:hidden to neutralise
any Unraid global h3 styles that could force card width
- network_watchdog: remove ISSUES++ for Tailscale — it's notify-only by
design, no restart possible; cascading to system watchdog exit 1 was noise
- docker_watchdog: move banner to top of cycle so body errors appear under
the correct section header, not visually inside Resource Manager output
- CPU/GPU cards: flex-column so body div takes remaining height
- CPU canvas: flex:1 + min-height:36px instead of fixed 36px
- GPU: add vvGpuHistory + canvas (purple, same style as CPU); canvas flex:1
stretches to fill whatever space the row's tallest card leaves over
- External IP: switch from DDNS domain lookup to curl ifconfig.me with 5-min
cache; DDNS approach broke because host detection uses HOST1_NAME but conf
has HOST1, so domain was never resolved
- Tailscale IP: switch from hardcoded tailscale0 iface to `tailscale ip -4`;
interface name varies (tailscale0, tailscale1) across kernels/installs
media.php: Emby/Jellyfin/Plex session poller for Streams card
confform.php: conf file form helpers for Config tab
script.php: per-script run/log API endpoint for Scheduler tab
vv_network_stats() calls vv_conf_vars() and vv_detect_host() which are defined in
config.php. monitor.php had no require for it, causing a PHP fatal error on every
poll → empty JSON → all monitor cards stuck on "Loading...".
Monitor tab:
- CPU card: per-core bars (freq-colored), overall usage bar, rolling history chart
- Memory card: breakdown by system/VM/ZFS/Docker/free with progress bars
- Network card: LAN + ext + Tailscale IP display, live RX/TX chart with auto-scale
- Streams card: Emby/Jellyfin/Plex sessions with playback bar and server badges
- GPU card: adds power draw, encode%, decode% meters
- Transcode card: ramdisk vs SSD location pill, session count, flip history
- Canvas chart helpers: vvDrawChart, vvDrawNetChart, vvMeter, vvFmtBps, vvFmtGib
api/monitor.php: adds cpu, mem, net to response payload
- Add .vv-save-check span to each job row (orchestrators + children)
- CSS: fade-in-then-out animation (visible 1.2s, fades over 2s)
- vvFlashSaved(job): triggers the checkmark animation on a specific row
- vvFlashStatus(el, msg, ok): shows timed status text (was called but missing)
- vvSaveJob: now shows per-job checkmark on successful save
- vvSaveAll: shows checkmark on each job as its response arrives, plus
global "✓ All saved" / "✗ Some failed" status text below the button
Unraid's system crond is dcron (dillon's cron daemon 4.5), which reads
/etc/cron.d/ for system crontabs WITHOUT a username field, and refuses
to execute files that are group- or world-writable.
vv_cron_rebuild() was writing "SCHED root bash ..." (with username field)
and leaving the file world-writable (666), so dcron silently ignored every
entry. Fixed:
- Remove CRON_USER from the generated cron line format
- chmod 0600 the cron file after writing it
- Change suggestion title color from bright green (#6fcf97) to soft grey-blue (#7a9eb5)
- Fix vv_parse_user_script_template() block-detection regex — $ end-anchor
failed on lines ending with multibyte U+2500 ─ chars; removed anchor so
parser now returns all 22 blocks from user_script_plug-in.sh
- Cron and Advanced button vertically aligned in their respective rows
- Verbose checkbox next to Log button — saves immediately, persists to schedule.json
- --log flag injected into cron lines, manual Run, and Dry Run when verbose enabled
- Coffee report rewritten as lean orchestrator over COFFEE_REPORT_SCRIPTS array
- COFFEE_REPORT_SCRIPTS added to master.conf (7 Sunday monitor scripts)
- Scheduler Advanced tab auto-discovers children via existing array detection
Gmer4Lfe