Operations
Last updated
Day-2 operator reference. For day-1 install see installation.md.
Standard daily / weekly cadence
| When | Action |
|---|---|
| Daily | Review /jabali-admin/security → CrowdSec → Decisions. Spot any allowlist additions you need to make. |
| Daily | Review /jabali-admin/notifications (or the bell) for cert.renew.fail, backup.fail and service.down. |
| Weekly | jabali update, pulls latest fixes; auto-applies migrations. |
| Weekly | Inspect /jabali-admin/server-status → trends. Spot disks filling up before quota alarms fire. |
| Monthly | Verify a restore round-trip from at least one backup destination. |
Common operator tasks
Lock out a panel user
jabali user suspend <id> --reason "…"
jabali user unsuspend <id>
(The same cascade as Suspend in the admin UI.)
Force-reset a user’s password / 2FA
jabali user password <email> # generates new password, prints once
jabali user 2fa-reset <email> # strip TOTP + recovery codes (CLI escape hatch)
Get an admin back in
jabali user 2fa-reset <email|username|user-id> # when the admin lost their 2FA device
jabali user password <email|username|user-id> --link
--link prints a one-click recovery URL, valid 24 hours.
Rotate the DB root password
/jabali-admin/settings → Database → Root Password → Rotate.
Or CLI:
jabali db root-password
Audited (success + failure), announced via notifications.
Move a domain between users
A first-class Change owner action reassigns a domain to a different user; the vhost, DNS, and mail follow the new owner on the next reconcile. The domain’s Python apps, Git repository settings and Composer project folder setting belong to the old owner and are deleted. A Docker app of the old owner is unlinked from the domain and keeps its data; a server-level one stays. The admin can also rename a tenant’s username in place — the panel re-prefixes that user’s MariaDB databases, DB users, and PostgreSQL shadow roles to match.
Add a new IPv4 to the pool
- Bring the IP up on the host (
ip addr add <ip>/<prefix> dev <iface>). /jabali-admin/ips→ Add.- Assign per-domain via Domains → Edit → Listen IP.
Verify the public installer URL (get.jabali-panel.com)
Short install URL. After a release or any DNS/redirect change, confirm it still serves the installer:
curl -fsSL https://get.jabali-panel.com | head -5 # expect the bash shebang + installer header
If it fails, the canonical fallback always works:
curl -fsSL https://github.com/shukiv/jabali-panel/raw/main/install.sh | sudo bash.
Setup + redirect config live in deploy/get-installer/.
Add a new PHP version
apt install php8.x-fpm php8.x-cli php8.x-mbstring …- The panel detects it on the next reconciler tick; appears in
/jabali-admin/php-pools. - Domains can pick it under Edit → PHP Version.
Common failures
| Symptom | First thing to check |
|---|---|
| Panel UI says “502 Bad Gateway” | systemctl status jabali-panel; journalctl -u jabali-panel -f |
| Cert never issues | jabali ssl list → see the row’s last error; journalctl -u jabali-agent -f while you toggle SSL again |
Dirty database version N on boot | A failed migration. jabali migrate up to retry; if it’s deterministic, file an issue with the journal output. |
nginx -t fails after a domain edit | Missing FastCGI cache keyzone (if you enabled cache before jabali update landed the keyzone conf). Run jabali update once. |
| Mail accounts work in IMAP but not SMTP submission | Bulwark may be down. systemctl status bulwark; check :587 listener. |
| User app suddenly can’t reach external API | Per-user egress firewall may have dropped the destination. Users → Edit → Egress. |
Repair
jabali repair --diagnose # default: report only
jabali repair --auto # safe auto-fixes (perms, drop-in regen, no destructive)
jabali repair --all --yes # destructive (wipes nginx tmp, regen pool sockets) — only when --diagnose tells you to
jabali repair covers 7 detectors, grouped by destructiveness.
Updates → Repair Center → Auto-Repair safe issues runs jabali repair --auto. That includes docroot-www-data-group, which puts docroot files back in group www-data and sets setgid on their directories, so nginx can serve newly uploaded media. A broken row that Auto-Repair leaves says why. A destructive one shows its jabali repair --<id> --yes command. One with no fix (dirty-migration, fpm-masters-down) needs your decision. In --diagnose output these carry [destructive] and [manual] after the label.
Migration cleanup
After a successful cPanel / DA / Hestia restore: jabali domain prune-orphans lists sites in nginx sites-enabled that have no panel row; jabali domain prune-orphans --apply deletes them.
--apply lists sites-enabled again after the deletes. Each site shows as deleted, or as FAILED with the reason when it is still there, and the command exits non-zero if any site is left. With --json, the result has deleted and failed (site and reason) lists. Updates → Domain repair → Delete orphans runs the same command and lists any site left, with its reason.
Diag bundle
/jabali-admin/support → encrypted bundle, mailto: send. See support.md.