How the watchdog restarts failed services automatically
On this page
The watchdog keeps the server's services running without you. Every minute it checks that each service is running and really answering. A service that has stopped or hangs is restarted, and you get an email. There is nothing to turn on.
What it checks
Every minute, for each service that is enabled on the server:
- the service must be running, and
- for the main services, it must answer a real request:
- web server — a page request on the server itself
- panel web server — the administrator panel on port 2087
- MariaDB and PostgreSQL — a database query
- DNS — a lookup
- Postfix — a greeting on port 25
- Dovecot — a greeting on port 143
The services watched include the web servers, PHP-FPM, MariaDB, PostgreSQL, DNS, Postfix, Dovecot, DKIM signing, the spam filter, Redis, FTP, the antivirus, fail2ban, cron, SSH, the agent and the panel's own helper services.
What it does when a service fails
- Stopped: the service is restarted at once.
- Running but not answering: usually a busy moment (a backup, a large query). The watchdog waits until the service has not answered for 3 minutes in a row, then stops it (forcing it if it hangs) and starts it again. Databases are never forced; they are only stopped cleanly.
- You get an email at the Contact email of Server Settings, with the service and what happened.
- After 3 restarts in one hour, the watchdog stops trying for that hour and tells you the service needs a look. This prevents a restart loop.
See what the watchdog did
- Open Service Manager in the menu (section Server).
- Look at the Watchdog table at the bottom. Each line shows when, which service and what happened:
- Restarted — the service was down and works again.
- Could not restart — the restart failed; the last lines of the service's log are shown.
- Failing repeatedly — not restarted any more this hour — it needs your attention.
The table shows the last 30 events. No service had to be restarted. means all is well.
Good to know
- The watchdog runs as its own system timer, separate from the agent, so it restarts services even when the agent is stopped. It also watches the agent (
sahaabd) itself. The email and the table are filled in by the agent, so they follow once the agent runs again. - The email reaches you only if the Contact email in Server Settings is a valid address.
Common questions
A service keeps being restarted. What should I do?
Read the detail in the Watchdog table, then the service's own log on the server, for example journalctl -u mariadb -n 50. Frequent restarts usually mean a full disk, too little memory or a broken configuration.
Can I turn the watchdog off?
There is no setting for it in the panel. It only acts on services that are enabled on the server.
Thank you — this helps us improve the article.
Still stuck?
Ask the AI about your exact situation — it answers from this knowledge base — or contact us.
Related articles
- How to restart a service (web, mail, DNS, database)
Check the status of the web server, mail, DNS, database and other services in the Service Manager and restart one with a click.
- Server settings explained: name, IP, nameservers, time zone
Every option on the Server Settings page: brand name, hostname, shared IP, contact email, nameservers, language, time zone, TTL and web server.
- The admin dashboard explained: what each number means
What the administrator dashboard shows: accounts, domains, tasks, server load, memory, disk, websites that are down and new versions.
- Tasks stay queued: the agent (sahaabd) is not running
Changes do not happen and the panel says tasks have been waiting over 2 minutes? Check and restart the sahaabd agent that carries out every task.