Skip to content

Critical bug: deleting a user crashes Hiddify Manager and makes the server unresponsive #5578

Description

@Fatmann666

Hello Hiddify development team,
I have found a critical issue in the latest Hiddify Manager update.

When I delete a user from the Hiddify Manager web panel, the system becomes completely unresponsive.
After clicking Delete user:

  1. The user deletion starts.
  2. Hiddify Manager stops responding.
  3. The web panel becomes inaccessible.
  4. The server itself becomes unresponsive.
  5. The problem cannot be recovered by restarting the Hiddify services.
  6. The only way to restore the server is to perform a full VPS reboot.
    This happens specifically when deleting a user through the Hiddify Manager panel.
    Environment
  • OS: Ubuntu 22.04.5 LTS
  • Hiddify Manager: 13.0.3
  • Server architecture: x86_64 / amd64
  • RAM: approximately 1 GB
  • Swap: 1 GB
  • Installation: self-hosted VPS
  • Server IP: 185.205.194.230
    The problem appeared after updating Hiddify Manager.
    Impact
    This is a critical issue because deleting a single user can effectively take down the entire VPS. A complete server reboot is required to recover.
    Could you please investigate this issue?
    In particular, it would be useful to check whether the user deletion process can cause:
  • an infinite loop;
  • excessive memory consumption;
  • a deadlock;
  • runaway CPU usage;
  • a problem in hiddify-core;
  • a database/SQLAlchemy issue;
  • or a process/resource leak during configuration regeneration.
    I can provide logs, configuration information, or reproduction steps if needed.
    Please let me know which logs you would like me to collect immediately after reproducing the problem.
    Thank you.
    Best regards,
    Hiddify user

Activity

  1. estaji commented on Sep 29, 2026

    @estaji

    Hello
    I have exactly the same problem. In my case, restarting services shows:

    Hiddify: Command restart
    ==========================================
    
    ----------------------------------------------------------------
    Name                           Old Status           New Status
    Error: Key not found: warp_mode
    hiddify-core                   active      --->     active
    hiddify-mysql                  active      --->     active
    hiddify-rpxy-l4                active      --->     active
    

    Some main service has wierd errors logs like:

    journalctl -e -u hiddify-panel-background-tasks.service
    Sep 29 11:32:27 xxx systemd[1]: hiddify-panel-background-tasks.service: Failed to op>
    Sep 29 11:32:27 xxx systemd[1]: hiddify-panel-background-tasks.service: Failed to op>
    
    
    journalctl -e -u hiddify-dnstm-router.service
    Sep 29 11:32:27 eu-server systemd[1]: hiddify-dnstm-router.service: Failed to open /etc/sy>
    Sep 29 11:32:27 eu-server systemd[1]: hiddify-dnstm-router.service: Failed to open /etc/sy>
    
    

    Any solution?

  2. hesamzakerirad commented on Oct 4, 2026

    @hesamzakerirad

    Same issue here. Whenever I update a user everything crashes and I see Error: Key not found: warp_mode error.

    I have to update the entire server using cli to fix the issue.

  3. 4ntoine commented on Oct 8, 2026

    @4ntoine

    Having the same issue, any solution?

  4. blshkv commented on Oct 9, 2026

    @blshkv

    I tracked this down on a v13.0.3 server. It's caused by scripts/migrate_layout.sh, and it's already fixed on dev (7ce3eab) but not yet in main/beta.

    Cause

    scripts/install.sh runs migrate_layout.sh on every call, including "Apply configs" and install.sh apply_users (what the panel runs when you change a user). Its disable_all_services searches /opt/hiddify-manager/ for *.service files. It skips services/ but not old/, where the upgrade moved the previous layout. The files in old/ have the same names as the current units (hiddify-haproxy, hiddify-nginx, hiddify-panel, hiddify-redis, hiddify-xray, mtproxy, ...). So it runs systemctl disable --now on the live services. That stops them and removes their symlinks from /etc/systemd/system.

    • A full install (CLI update, or the @reboot reinstall) recreates the units in each service's install.sh, so everything comes back. That's why updating from the CLI "fixes" it.
    • Apply configs / apply users run with DO_NOT_INSTALL=true, so nothing recreates haproxy, nginx, panel and redis. They stay gone, and hiddify restart only lists core/mysql/rpxy-l4. The Failed to open /etc/systemd/system/... messages in the journal are the deleted symlinks.

    To check whether you're affected:

    find /opt/hiddify-manager/ -type d -name services -prune -o -type f -name '*.service' -print

    If this prints anything under old/, you're hit by this.

    Fix

    Make the same change as dev in /opt/hiddify-manager/scripts/migrate_layout.sh (also skip old/):

    sudo sed -i 's#-type d -name "services" -prune -o -type f#-type d -name "services" -prune -o -type d -name "old" -prune -o -type f#' /opt/hiddify-manager/scripts/migrate_layout.sh

    Then run a full install once to restore the deleted units:

    sudo bash /opt/hiddify-manager/scripts/install.sh --no-gui

    After the change, I ran apply_configs.sh and install.sh apply_users, and all units (haproxy, nginx, panel, redis, xray, core, mysql, rpxy-l4, mtproxy) stayed loaded and running. Before the change, apply_configs.sh removed haproxy, nginx, panel and redis every time.

    Note: the next update unpacks a new release over the patched file and will overwrite it until the fix is in a stable release. If you don't need anything in /opt/hiddify-manager/old/, deleting it also prevents the problem.

    Error: Key not found: warp_mode is unrelated and harmless. It's just a missing config key that restart.sh reads.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions