Skip to content

Operations ​

Day-to-day running of a Vexlyx server: watching resource usage, backing up and restoring, the firewall, restarting the mail/DNS/database stack, reclaiming disk space, managing who has access, and upgrading. Most of this is ADMIN-only.

Monitoring ​

Monitoring shows live CPU/RAM/disk gauges (pushed every 5 seconds), uptime and load average, a historical trend chart (1h/24h/7d/30d), and a per-container resource table. A badge and toast appear automatically if CPU passes 80%, RAM 85%, or disk 90% — the same numbers the dashboard home page's Server Health widget summarizes. History is sampled and stored every 60 seconds, so the trend chart survives a panel restart.

Backups ​

Backups takes full-system snapshots — every project's source tree, every database, every mail domain's Maildir, and every domain's DNS records — bundled into one archive per snapshot.

  • Schedule: one cron schedule for the whole server, daily at 3am by default. Change it in the settings card; it takes effect immediately, no restart needed.
  • Retention: keeps the most recent N daily snapshots plus M older ones spaced about a week apart (7/4 by default), deleting the rest automatically after each run.
  • Restore is per-item, not all-or-nothing: open a snapshot and pick a single project, database, mail domain, or DNS zone to restore in place — there's no single "restore everything" button. Restoring is destructive to that one item, so it's behind a confirmation dialog.
  • Manual backups: click Backup Now any time; progress streams live.

Not included in a snapshot: SSL certificates, vacation-responder settings, SFTP accounts, and project environment variables — worth knowing before you rely on a restore to bring a project fully back. Backups are stored on local disk only; there's no built-in off-server/S3 copy, so for real disaster recovery you should periodically copy the backups directory somewhere else yourself.

Firewall ​

Firewall manages the server's real UFW firewall — not a simulated view. Add or remove rules (port, protocol, optional source IP/CIDR, allow/deny, comment), and set the default incoming/outgoing policy. Rules created by the installer itself (SSH, HTTP/HTTPS, mail ports) show up too, marked as system-managed and read-only.

Vexlyx refuses any change that would lock you out — you can't deny or delete the rule protecting SSH or the panel's own API port, and you can't set a deny-by-default incoming policy until at least one of those ports has an explicit allow rule. If a change is rejected, that's why.

Service status ​

Services (ADMIN-only) shows live status for the five containers the hosting stack depends on — Postfix, Dovecot, CoreDNS, PostgreSQL, and Redis — plus a read-only Docker daemon reachability check. Each service can be started, stopped, or restarted (stop/restart ask for confirmation first, since these are shared infrastructure, not one user's project), and you can pull a recent log tail on demand without shelling into the server.

If mail or DNS stops working, this is the first place to check whether the underlying container is actually running before digging further.

Reclaiming disk space ​

Every redeploy used to leave the previous build's image on disk forever. The Monitoring page's Docker Disk Usage card breaks down space used by images, containers, volumes, and build cache, with a manual Clean up button (removes unused containers and images; volumes and build cache are shown but never auto-removed). Two things can run this automatically instead of manually:

  • Scheduled cleanup — an opt-in cron schedule, off by default.
  • Prune after redeploy — an opt-in setting that removes a project's previous image right after its new one is confirmed running, so redeploying repeatedly doesn't slowly fill the disk.

Every cleanup run (manual, scheduled, or redeploy-triggered) is logged with what was removed and how much space came back.

Users, roles, and permissions ​

Three roles:

  • ADMIN — everything, including user management, firewall, and backups.
  • RESELLER — creates and manages their own sub-accounts (plain USER accounts), capped by a maxSubAccounts quota, and sees only themselves plus their sub-accounts on the Users page.
  • USER — manages their own projects, domains, databases, and mailboxes, each capped by their own resource quotas (projects/domains/databases/mailboxes — null means unlimited).

An ADMIN sets quotas and roles from Users. A role change or quota update takes effect on that user's very next request — no re-login required.

If a USER needs one extra capability without a full promotion to RESELLER — say, managing DNS — grant it individually from the same edit dialog: fine-grained permissions (canManageDns, canManageFirewall, canManageBackups, canCreateSubAccounts) are additive on top of whatever the role already allows, and can be revoked the same way.

By default, a fresh install doesn't allow public self-registration — the first admin account (from the install summary) creates every other account through Users.

Audit log ​

Audit Log (ADMIN-only) records who changed what: user role/quota/permission changes, user create/delete, firewall rule and policy changes, backup restore/delete, and create/delete for projects, domains, databases, and mailboxes. Every entry keeps a before/after snapshot of exactly what changed — filterable by actor, action, and date range. Automated actions (like a scheduled backup's own retention cleanup) intentionally don't appear here, since they're not something a person did.

Upgrading ​

To upgrade Vexlyx to the latest release — or to repair a broken install — just re-run the installer on the server:

bash
curl -fsSL https://vexlyx.atlantiqs.org/install.sh | bash

The installer installs the newest tagged release (for example v0.2.0), not whatever happens to be on the development branch, and re-running it moves the install to the next release when one is published. To pin a version, or to follow development instead, set VEXLYX_REPO_REF:

bash
curl -fsSL https://vexlyx.atlantiqs.org/install.sh | sudo VEXLYX_REPO_REF=v0.2.0 bash   # exact release
curl -fsSL https://vexlyx.atlantiqs.org/install.sh | sudo VEXLYX_REPO_REF=main bash     # development

An install that already follows a branch keeps following it until you set VEXLYX_REPO_REF=latest.

Administrators see a notice in the dashboard when a newer release exists (v0.4.3 and later), with links to its release notes and to this section. The check only reads GitHub release metadata — it never downloads or installs anything — and its result is cached for six hours. On a server that cannot reach api.github.com the notice simply stays hidden; set UPDATE_CHECK_ENABLED=false in /etc/vexlyx/vexlyx.env to skip the lookup entirely.

When a release changes Traefik's static configuration (for example the upload timeouts added in v0.4.1), the installer restarts the proxy once, and only then. Every site on the server is unreachable for a few seconds during that restart; an upgrade that does not change the configuration never restarts it.

Some releases recreate more of the stack. v0.4.2 gives Redis a password and moves projects onto their own network (see Network layout). The installer generates REDIS_PASSWORD in /etc/vexlyx/vexlyx.env, recreates Redis, Traefik, the database servers, the API and the dashboard (every site pauses for a few seconds), and reloads the nginx container of each PHP or WordPress project once it has moved (those sites can pause briefly once more). Logins and queued jobs are kept. If you connect to Redis yourself, use the password from that file.

Servers installed before v0.3.0

Older versions committed a mailbox file (docker/dovecot/config/users) containing accounts with a publicly known password, and the panel keeps rewriting that file on your server. Git therefore refuses to update such a checkout, and the installer leaves a checkout with local changes alone, so re-running the installer does not upgrade it. Follow the one-time procedure in Remove exposed development mail credentials first (it takes a backup, updates the checkout, and removes the known credentials; Dovecot refuses to start in production while they are present). Servers installed from v0.3.0 on are not affected.

Every installer step checks real state before acting, so re-running it is always safe: nothing that already succeeded is redone, secrets are never regenerated or rotated, and only what actually needs updating (code, container images, migrations) changes. See the installer reference for exactly what each step does.

Next steps ​

  • Email — the mail stack has its own operational tools (queue, delivery log, DKIM rotation) beyond what's covered here.
  • Domains, subdomains and SSL — SSL certificates renew automatically, but the domain's SSL tab is where you'd check on that.

Released under the MIT license.