Telegram Bot poor performance

[MihailRobot]

Telegram bot seems to be written really really poorly. It takes 30s - 1m to respond to any message because it writes to disk EVERY command and /mp_secrets fetches metrics once, then calls get_user_stats_tg (which re-fetches and re-parses) for every single secret etc. There are a lot of problems like that

[SamNet-dev]

Fixed. Three changes:

1. Long-polling (instant response)
The bot was using timeout=1 with sleep 30 between polls — average 15s delay. Now uses timeout=25 long-poll so Telegram holds the connection open until a message arrives, then responds immediately.

2. Single awk pass for metrics
/mp_secrets was running 3 separate awk calls per user (N users = 3N subprocesses). Now uses one awk pass for all users. Same fix applied to update_traffic (runs every 60s) and get_stats.

3. Eliminated unnecessary subshells
/mp_status, /mp_traffic, periodic report — replaced echo "$x"|awk '{print $N}' patterns with read -r which is a bash builtin (no subprocess).

Run mtproxymax update then restart the Telegram service:

systemctl restart mtproxymax-telegram

[MihailRobot]

Nice to see that you are checking issues but you didnt fix the biggest problem. Right now script creates temp files for each telegram message which is a really big hit on performance. You can easily create no temp files by doing something like this:

tg_send() {
    local msg="$1"
    local label=$(get_setting "TELEGRAM_SERVER_LABEL")
    label="${label:-MTProxyMax}"
    
    local ip; ip=$(get_cached_ip)
    if [ -n "$ip" ]; then
        msg="[$(_esc "$label") | ${ip}] ${msg}"
    else
        msg="[$(_esc "$label")] ${msg}"
    fi
    

    curl -s --max-time 5 --connect-timeout 2 \
        "https://api.telegram.org/bot$(get_setting "TELEGRAM_BOT_TOKEN")/sendMessage" \
        --data-urlencode "chat_id=$(get_setting "TELEGRAM_CHAT_ID")" \
        --data-urlencode "text=$(printf '%b' "$msg")" \
        --data-urlencode "parse_mode=Markdown" >/dev/null 2>&1
}

tg_send_photo() {
    local photo="$1" caption="$2"
    local label=$(get_setting "TELEGRAM_SERVER_LABEL")
    label="${label:-MTProxyMax}"
    caption="[$(_esc "$label")] ${caption}"
    

    curl -s --max-time 10 --connect-timeout 2 \
        "https://api.telegram.org/bot$(get_setting "TELEGRAM_BOT_TOKEN")/sendPhoto" \
        -F "chat_id=$(get_setting "TELEGRAM_CHAT_ID")" \
        -F "photo=${photo}" \
        -F "caption=${caption}" \
        -F "parse_mode=Markdown" >/dev/null 2>&1
}

send_proxy_qr() {
    local ip="$1" port="$2" secret="$3" caption="${4:-Scan in Telegram to connect}"
    local hl="https://t.me/proxy?server=${ip}&port=${port}&secret=${secret}"
    local el=$(printf '%s' "$hl" | sed 's/:/%3A/g; s|/|%2F|g; s/?/%3F/g; s/=/%3D/g; s/&/%26/g')
    tg_send_photo "https://api.qrserver.com/v1/create-qr-code/?size=300x300&data=${el}" "$caption"
}

[SamNet-dev]

@MihailRobot Good point on the temp files. Fixed — replaced all mktemp + rm patterns with bash process substitution (-K <(printf ...)). Zero disk I/O, token still not visible in process args.

The suggested approach of putting the token directly in the curl URL would expose it in ps aux / /proc/*/cmdline, which is why we use -K instead. Process substitution gives both: no temp files AND no token in the process list.

Run mtproxymax update && systemctl restart mtproxymax-telegram.

[MihailRobot]

Nice this update fixed exactly all issues i was having. Thank you!