Repair and uninstall

The vmstack-panel command on the panel server: status, repair, tune after a resize, certificates, new sign-in links, database dumps and uninstalling.

The installer leaves a management command on the panel server, vmstack-panel. Run it as root. Every release brings its own copy, so it is always the version that matches the panel.

Command What it does
vmstack-panel status The services, the version, the certificate, PHP-FPM's status and a worker check on every queue
vmstack-panel repair Writes every configuration file again and restarts what changed
vmstack-panel tune Computes the tuning from CPU and memory again and applies it
vmstack-panel cert [--staging|--production] [--file fullchain.pem --key key.pem] Gets a Let's Encrypt certificate again, or installs your own
vmstack-panel admin-link <email> A new one-time sign-in link for an administrator who has not set a password
vmstack-panel backup Dumps the database now (a timer does it daily)
vmstack-panel update [--version X] [--channel C] [--check] Updates the panel; see Panel updates
vmstack-panel uninstall [--purge] Removes the panel, after a final backup

Check the panel#

vmstack-panel status

It lists each unit (nginx, PHP-FPM, PostgreSQL, Redis, vmstack-scheduler, vmstack-relay and every vmstack-worker@…) with its state, the certificate's type and end date, and then puts a heartbeat on every queue and waits up to 15 seconds for a worker to take it, naming any queue no worker took. The same information is in the panel under System › Health.

To restart everything of the panel:

systemctl restart vmstack-panel.target

Repair#

vmstack-panel repair

repair writes again every file the installer manages (nginx, PHP-FPM and PHP settings, PostgreSQL and Redis settings, systemd units, the firewall, the tuning files), checks the links and permissions of /opt/vmstack, and restarts only what changed. It prints the files it rewrote, or "Nothing changed: every file was as it should be." Run a second time, it changes nothing.

It never touches the secrets in .env, the database's data or the certificate. bash install.sh --repair does the same.

Use it when a configuration file was edited by hand or damaged, or after changing a value in /etc/vmstack/panel.conf.

Your own settings#

Every file the installer writes starts with a line saying VMStack manages it and that tune and repair rewrite it. Put your own changes in the files read after the installer's, which it never touches:

Program Your file
nginx /etc/nginx/conf.d/vmstack-local.conf
PHP-FPM pool /etc/php/8.5/fpm/pool.d/zz-local.conf
PHP (web, command line) /etc/php/8.5/fpm/conf.d/99-local.ini, /etc/php/8.5/cli/conf.d/99-local.ini
PostgreSQL /etc/postgresql/<version>/main/conf.d/zz-local.conf
Redis /etc/redis/vmstack-local.conf
Kernel /etc/sysctl.d/99-local.conf

Tune after a resize#

vmstack-panel tune

Run it after you give the server more (or fewer) CPUs or memory. It recomputes PHP-FPM's workers, opcache, PostgreSQL's memory and connections, Redis's memory and the swap file, applies them and restarts what changed. Once PHP-FPM has run for an hour, it uses the measured memory of a PHP worker instead of an estimate. It prints the values it chose.

To pin a value, set it in /etc/vmstack/panel.conf (see Installer options) and run tune.

Certificates#

The Let's Encrypt certificate renews by itself (certbot's timer; nginx is reloaded after each renewal).

To get it again, for example after a failed install left a self-signed one:

vmstack-panel cert

When Let's Encrypt refuses, the command prints its reason and the panel keeps a self-signed certificate. Usual causes: the domain does not point at the server yet, or port 80 is not reachable from the internet.

--staging switches to Let's Encrypt's staging certificates (tests), --production back.

To use a certificate of your own:

vmstack-panel cert --file /root/fullchain.pem --key /root/key.pem

Both must be PEM files. They are copied to /etc/vmstack/tls/, and certbot's renewal for the domain is turned off. Renewing your own certificate is then up to you: run the command again with the new files.

Database dumps#

A timer dumps the database every night at 02:30 into /opt/vmstack/shared/backups/daily-<date>.dump, keeping the last 7. vmstack-panel backup makes one now. These stay on the panel's server: copy them elsewhere with .env, or use Panel backup, which encrypts and sends them off the server.

Uninstall#

vmstack-panel uninstall

It asks you to type the panel's domain; anything else stops it with nothing removed (--confirm panel.example.com answers it in a script). Then:

  1. A final backup: the database, .env and storage/app into /root/vmstack-panel-<date>-<time>.tar.gz (mode 0600). It holds APP_KEY: keep it safe.
  2. Stops and removes the panel's units and timers, its nginx site (nginx's original configuration is put back and nginx is stopped, so ports 80 and 443 are free), its PHP-FPM pool (the distribution's default pool is put back), its firewall table, its sysctl, PostgreSQL and Redis settings, the swap file it made, the releases and the vmstack-panel command.
  3. Keeps the database, /opt/vmstack/shared (with .env) and /etc/vmstack/panel.conf, so a new install on the same server finds the data again.

With --purge, it also drops the database and its role and removes /opt/vmstack and /etc/vmstack.

The packages (nginx, PostgreSQL, Redis, PHP) are never removed; apt purge them if you want them gone.

Warning: Nodes are not told. They keep running their servers, and they keep trusting this panel's certificate authority. To manage them again, restore the panel from a backup (see Panel backup and restore); a new, empty panel cannot take them over without enrolling them again.

Installing again on the same server#

After an uninstall without --purge, run the installer again. It finds the database and .env, and the administrator keeps the password it had (the summary says so). An administrator who never set a password gets a new sign-in link.