moirk/Unraid_Intel-ARC_Deployment
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
0580c5e37e7106240217dea80ffade2338875d5c
Unraid_Intel-ARC_Deployment/tdarr_node_scaling.sh
Admin9705 0580c5e37e Update tdarr_node_scaling.sh
2025-03-16 17:34:31 -04:00

501 lines
18 KiB
Bash
Raw Blame History

#!/bin/bash
###################################
# Configuration
###################################
# Where your Tdarr Node log is located
TDARR_NODE_LOG_PATH="/mnt/user/appdata/n1-1/logs/Tdarr_Node_Log.txt"
# ------------- T1 (Required) -------------
T1_TAUTULLI_API_KEY="ebdb8c80fc2b461ea182243dbc1b27a1"
T1_TAUTULLI_URL="http://10.0.0.10:8181/api/v2"
####################################### IMPORTANT NOTE #######################################
# Additional Tautulli configurations are only needed if you run multiple Plex servers
# on the same GPU (e.g., a primary and a backup Plex instance). This lets the script
# track transcodes across all your servers sharing that single GPU.
#
# In most cases (99% of setups), you only have one Plex server (per gpu), so you can ignore
# any extra Tautulli entries.
# ------------- T2 (Optional) -------------
T2_TAUTULLI_API_KEY=""
T2_TAUTULLI_URL=""
# ------------- T3 (Optional) -------------
T3_TAUTULLI_API_KEY=""
T3_TAUTULLI_URL=""
# ------------- T4 (Optional) -------------
T4_TAUTULLI_API_KEY=""
T4_TAUTULLI_URL=""
###############################################################################################
# ----------- Tdarr Settings -------------
TDARR_ALTER_WORKERS=true # If true, we adjust GPU workers; otherwise we kill container on threshold
TDARR_DEFAULT_LIMIT=5 # Default GPU workers when watchers=0
TDARR_API_URL="http://10.0.0.10:8265" # WITHOUT /api/v2
CONTAINER_NAME="N1" # Name of your Tdarr Node Docker container
# ----------- Offset Setting ------------- IF >>> TDARR_ALTER_WORKERS=true
# We only start reducing workers when watchers >= OFFSET_THRESHOLD.
# e.g. If OFFSET_THRESHOLD=3, watchers <3 => no reduction,
# watchers=3 => reduce by 1, watchers=4 => reduce by 2, etc.
OFFSET_THRESHOLD=2 # For Tdarr Scaling, number of transcodes reached to start reducing GPU Workers (when TDARR_ALTER_WORKERS=true)
# ----------- Other -------------
WAIT_SECONDS=10 # Sleep after adjustments
BASIC_CHECK=3 # Poll interval (seconds) when idle
TRANSCODE_THRESHOLD=4 # For Tdarr Killer, number of transcodes reached to kill the tdarr node (when TDARR_ALTER_WORKERS=false)
###################################
# End of configuration
###################################
# Simple logging function with timestamp
log_message() {
echo "$(date '+%Y-%m-%d %H:%M:%S') - $1"
}
# ------------------------------------------------------------
# Function: find_latest_node_id
# ------------------------------------------------------------
find_latest_node_id() {
if [ -f "$TDARR_NODE_LOG_PATH" ]; then
local found
found=$(grep -oP '"nodeID":\s*"\K[^"]+' "$TDARR_NODE_LOG_PATH" | tail -n1)
echo "$found"
else
echo ""
fi
}
# Initialize global variable
TDARR_NODE_ID=""
# ------------------------------------------------------------
# Function: ensure_node_id_loaded
# ------------------------------------------------------------
ensure_node_id_loaded() {
if [ -n "$TDARR_NODE_ID" ]; then
return 0
fi
log_message "Attempting to retrieve nodeID from $TDARR_NODE_LOG_PATH"
local found
found=$(find_latest_node_id)
if [ -z "$found" ]; then
log_message "ERROR: Could not find any nodeID in $TDARR_NODE_LOG_PATH."
return 1
fi
TDARR_NODE_ID="$found"
log_message "Found nodeID: $TDARR_NODE_ID"
return 0
}
# ------------------------------------------------------------
# Function: refresh_node_id_if_changed
# ------------------------------------------------------------
refresh_node_id_if_changed() {
local latest
latest=$(find_latest_node_id)
if [ -z "$latest" ]; then
log_message "WARNING: Could not find any 'nodeID' lines in the log to refresh."
return
fi
if [ "$latest" != "$TDARR_NODE_ID" ]; then
log_message "NOTICE: nodeID changed from [$TDARR_NODE_ID] -> [$latest]. Updating."
TDARR_NODE_ID="$latest"
else
log_message "NOTICE: nodeID is still the same [$TDARR_NODE_ID]."
fi
}
# ------------------------------------------------------------
# Function: check_single_tautulli_connection
# ------------------------------------------------------------
check_single_tautulli_connection() {
local api_key="$1"
local url="$2"
if [ -z "$api_key" ] || [ -z "$url" ]; then
return 2
fi
log_message "Checking Tautulli at: $url"
local response
response=$(curl -s "${url}?apikey=${api_key}&cmd=get_activity")
if echo "$response" | jq . >/dev/null 2>&1; then
log_message "Tautulli OK: $url"
return 0
else
log_message "WARNING: Could not connect or invalid JSON: $url"
return 1
fi
}
# ------------------------------------------------------------
# Function: check_tautulli_connections_on_startup
# ------------------------------------------------------------
check_tautulli_connections_on_startup() {
# T1 must work
check_single_tautulli_connection "$T1_TAUTULLI_API_KEY" "$T1_TAUTULLI_URL"
if [ $? -ne 0 ]; then
log_message "ERROR: T1 not reachable. Exiting."
exit 1
fi
# T2..T4 are optional
check_single_tautulli_connection "$T2_TAUTULLI_API_KEY" "$T2_TAUTULLI_URL" || true
check_single_tautulli_connection "$T3_TAUTULLI_API_KEY" "$T3_TAUTULLI_URL" || true
check_single_tautulli_connection "$T4_TAUTULLI_API_KEY" "$T4_TAUTULLI_URL" || true
}
# ------------------------------------------------------------
# Function: fetch_transcode_counts_from_tautulli
# ------------------------------------------------------------
fetch_transcode_counts_from_tautulli() {
local api_key="$1"
local url="$2"
if [ -z "$api_key" ] || [ -z "$url" ]; then
echo "0 0"
return
fi
local resp
resp=$(curl -s "${url}?apikey=${api_key}&cmd=get_activity")
if ! echo "$resp" | jq . >/dev/null 2>&1; then
echo "0 0"
return
fi
local local_cnt remote_cnt
# Only count sessions that are transcoding video, not just audio
local_cnt=$(echo "$resp" | jq '[.response.data.sessions[]?
| select(.transcode_decision == "transcode" and .video_decision == "transcode" and (.ip_address | startswith("10.0.0.")))] | length')
remote_cnt=$(echo "$resp" | jq '[.response.data.sessions[]?
| select(.transcode_decision == "transcode" and .video_decision == "transcode" and (.ip_address | startswith("10.0.0.") | not))] | length')
echo "$local_cnt $remote_cnt"
}
# ------------------------------------------------------------
# Global watchers total
# ------------------------------------------------------------
total_count=0
# ------------------------------------------------------------
# Function: is_plex_transcoding_over_threshold
# ------------------------------------------------------------
is_plex_transcoding_over_threshold() {
log_message "Checking Plex transcodes..."
local total_local=0
local total_remote=0
# T1
read t1_local t1_remote < <(fetch_transcode_counts_from_tautulli "$T1_TAUTULLI_API_KEY" "$T1_TAUTULLI_URL")
total_local=$(( total_local + t1_local ))
total_remote=$(( total_remote + t1_remote ))
# T2
read t2_local t2_remote < <(fetch_transcode_counts_from_tautulli "$T2_TAUTULLI_API_KEY" "$T2_TAUTULLI_URL")
total_local=$(( total_local + t2_local ))
total_remote=$(( total_remote + t2_remote ))
# T3
read t3_local t3_remote < <(fetch_transcode_counts_from_tautulli "$T3_TAUTULLI_API_KEY" "$T3_TAUTULLI_URL")
total_local=$(( total_local + t3_local ))
total_remote=$(( total_remote + t3_remote ))
# T4
read t4_local t4_remote < <(fetch_transcode_counts_from_tautulli "$T4_TAUTULLI_API_KEY" "$T4_TAUTULLI_URL")
total_local=$(( total_local + t4_local ))
total_remote=$(( total_remote + t4_remote ))
total_count=$(( total_local + total_remote ))
log_message "Found $total_local local & $total_remote remote => total=$total_count, threshold=$TRANSCODE_THRESHOLD"
# Return 0 if watchers >= threshold
if [ "$total_count" -ge "$TRANSCODE_THRESHOLD" ]; then
return 0
fi
return 1
}
# ------------------------------------------------------------
# Function: is_container_running
# ------------------------------------------------------------
is_container_running() {
local s
s=$(docker inspect -f '{{.State.Running}}' "$CONTAINER_NAME" 2>/dev/null)
[ "$s" = "true" ]
}
# ------------------------------------------------------------
# Function: adjust_tdarr_workers
#
# We only start subtracting from TDARR_DEFAULT_LIMIT once watchers >= OFFSET_THRESHOLD.
# Example with OFFSET_THRESHOLD=3:
# watchers=2 => no reduce
# watchers=3 => reduce by 1
# watchers=4 => reduce by 2, etc.
# ------------------------------------------------------------
adjust_tdarr_workers() {
ensure_node_id_loaded || return
local watchers="$1"
# 1) Calculate how many watchers are above the offset
# If OFFSET_THRESHOLD=0, use watchers directly
# If OFFSET_THRESHOLD>0 and watchers < OFFSET_THRESHOLD => watchersOverOffset=0
# If OFFSET_THRESHOLD>0 and watchers=3 => watchersOverOffset=1 => reduce by 1
# If OFFSET_THRESHOLD>0 and watchers=4 => watchersOverOffset=2 => reduce by 2, etc.
local watchersOverOffset
if [ "$OFFSET_THRESHOLD" -eq 0 ]; then
# When offset is 0, reduce by exactly the watcher count
watchersOverOffset=$watchers
else
# When offset >0, start reducing only after reaching threshold
watchersOverOffset=$(( watchers - OFFSET_THRESHOLD + 1 ))
if [ "$watchersOverOffset" -lt 0 ]; then
watchersOverOffset=0
fi
fi
# 2) Desired = TDARR_DEFAULT_LIMIT - watchersOverOffset
local desired=$(( TDARR_DEFAULT_LIMIT - watchersOverOffset ))
if [ "$desired" -lt 0 ]; then
desired=0
fi
log_message "watchers=$watchers => watchersOverOffset=$watchersOverOffset => desiredWorkers=$desired"
# poll-worker-limits
local poll_resp
poll_resp=$(curl -s -X POST "${TDARR_API_URL}/api/v2/poll-worker-limits" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'"}}')
local current
current=$(echo "$poll_resp" | jq '.workerLimits.transcodegpu' 2>/dev/null)
if [ -z "$current" ] || [ "$current" = "null" ]; then
log_message "ERROR: Could not retrieve current GPU worker limit for nodeID='$TDARR_NODE_ID'. Will re-check log for a new ID."
refresh_node_id_if_changed
return
fi
log_message "Current GPU worker limit: $current"
local diff=$(( desired - current ))
if [ "$diff" -eq 0 ]; then
log_message "Already at the desired GPU worker limit ($desired)."
return
fi
local step
if [ "$diff" -gt 0 ]; then
step="increase"
log_message "Need to increase by $diff"
else
step="decrease"
diff=$(( -diff ))
log_message "Need to decrease by $diff"
fi
local i=0
while [ $i -lt $diff ]; do
curl -s -X POST "${TDARR_API_URL}/api/v2/alter-worker-limit" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'","process":"'"$step"'","workerType":"transcodegpu"}}' \
>/dev/null 2>&1
i=$(( i + 1 ))
sleep 1
done
log_message "GPU worker limit adjustment complete."
}
# ------------------------------------------------------------
# Main Script
# ------------------------------------------------------------
ensure_node_id_loaded
check_tautulli_connections_on_startup
# Main loop with protection against duplicate operations
last_operation=""
last_gpu_limit=0
consecutive_duplicates=0
# Set initial GPU workers on startup
if [ "$TDARR_ALTER_WORKERS" = "true" ]; then
log_message "Setting initial GPU workers to default limit: $TDARR_DEFAULT_LIMIT on startup"
# Ensure we have nodeID before trying to set workers
ensure_node_id_loaded || {
log_message "ERROR: Could not get nodeID, can't set initial GPU workers"
sleep 5 # Wait a bit and continue, will try again in the main loop
}
if [ -n "$TDARR_NODE_ID" ]; then
# Get current limit
current_limit=$(curl -s -X POST "${TDARR_API_URL}/api/v2/poll-worker-limits" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'"}}' | \
jq '.workerLimits.transcodegpu' 2>/dev/null)
if [ -n "$current_limit" ] && [ "$current_limit" != "null" ]; then
# Calculate how many workers to add/remove
diff=$(( TDARR_DEFAULT_LIMIT - current_limit ))
if [ "$diff" -ne 0 ]; then
step=""
count=0
if [ "$diff" -gt 0 ]; then
step="increase"
count=$diff
log_message "Need to increase by $diff to reach default limit"
else
step="decrease"
count=$(( -diff ))
log_message "Need to decrease by $(( -diff )) to reach default limit"
fi
i=0
while [ $i -lt $count ]; do
curl -s -X POST "${TDARR_API_URL}/api/v2/alter-worker-limit" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'","process":"'"$step"'","workerType":"transcodegpu"}}' \
>/dev/null 2>&1
i=$(( i + 1 ))
sleep 1
done
log_message "Initial GPU worker limit set to $TDARR_DEFAULT_LIMIT"
else
log_message "GPU workers already at desired default limit: $current_limit"
fi
else
log_message "ERROR: Could not get current GPU worker limit"
fi
fi
fi
while true; do
if is_plex_transcoding_over_threshold; then
# watchers >= threshold
if [ "$TDARR_ALTER_WORKERS" = "true" ]; then
# Check if we're doing the same operation repeatedly
operation="reduce_workers_$total_count"
# Get current limit to check if it changed
current_limit=$(curl -s -X POST "${TDARR_API_URL}/api/v2/poll-worker-limits" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'"}}' | \
jq '.workerLimits.transcodegpu' 2>/dev/null)
if [ "$operation" = "$last_operation" ] && [ "$current_limit" = "$last_gpu_limit" ]; then
consecutive_duplicates=$((consecutive_duplicates + 1))
if [ $consecutive_duplicates -gt 2 ]; then
log_message "Skipping duplicate worker adjustment (done $consecutive_duplicates times already)"
sleep "$WAIT_SECONDS"
continue
fi
else
consecutive_duplicates=0
fi
last_operation="$operation"
last_gpu_limit="$current_limit"
log_message "Threshold exceeded. Reducing GPU workers."
adjust_tdarr_workers "$total_count"
sleep "$WAIT_SECONDS"
else
# kill container
operation="kill_container"
if [ "$operation" = "$last_operation" ]; then
consecutive_duplicates=$((consecutive_duplicates + 1))
if [ $consecutive_duplicates -gt 2 ]; then
log_message "Skipping duplicate container management (done $consecutive_duplicates times already)"
sleep "$WAIT_SECONDS"
continue
fi
else
consecutive_duplicates=0
fi
last_operation="$operation"
if is_container_running; then
log_message "Threshold exceeded: Killing $CONTAINER_NAME"
docker kill "$CONTAINER_NAME"
else
log_message "$CONTAINER_NAME is already stopped."
fi
sleep "$WAIT_SECONDS"
fi
else
# watchers < threshold
if [ "$TDARR_ALTER_WORKERS" = "true" ]; then
# Check if we're doing the same operation repeatedly
operation="adjust_workers_$total_count"
# Get current limit to check if it changed
current_limit=$(curl -s -X POST "${TDARR_API_URL}/api/v2/poll-worker-limits" \
-H "Content-Type: application/json" \
-d '{"data":{"nodeID":"'"$TDARR_NODE_ID"'"}}' | \
jq '.workerLimits.transcodegpu' 2>/dev/null)
if [ "$operation" = "$last_operation" ] && [ "$current_limit" = "$last_gpu_limit" ]; then
consecutive_duplicates=$((consecutive_duplicates + 1))
if [ $consecutive_duplicates -gt 2 ]; then
log_message "Skipping duplicate worker adjustment (done $consecutive_duplicates times already)"
sleep "$BASIC_CHECK"
continue
fi
else
consecutive_duplicates=0
fi
last_operation="$operation"
last_gpu_limit="$current_limit"
adjust_tdarr_workers "$total_count"
fi
# Start container if needed
operation="start_container"
if [ "$operation" = "$last_operation" ] && is_container_running; then
consecutive_duplicates=$((consecutive_duplicates + 1))
if [ $consecutive_duplicates -gt 2 ]; then
log_message "Skipping duplicate container check (done $consecutive_duplicates times already)"
sleep "$BASIC_CHECK"
continue
fi
else
consecutive_duplicates=0
fi
last_operation="$operation"
if ! is_container_running; then
log_message "Below threshold -> Starting container $CONTAINER_NAME."
docker start "$CONTAINER_NAME"
else
log_message "Container $CONTAINER_NAME is already running."
fi
sleep "$BASIC_CHECK"
fi
done
Reference in New Issue View Git Blame Copy Permalink