Uplance Monitor types MySQL
Monitor whether MySQL accepts connections
Uplance opens a real connection to MySQL or MariaDB, authenticates, and runs a real query — the same three things your application has to do.
Free for 10 monitors · no credit card
The problem
“Too many connections” is not a crash. The process is running, the port is open, and every new connection is refused while the old ones sit idle in sleep.
Disk-full, a stalled replica and a table waiting on a lock all look the same from outside: healthy. Your users find them first.
What the check looks at
A real connection
MySQL or MariaDB, authenticated with the credentials you give it, exactly as your application authenticates.
A real query
A trivial query runs and has to return, so a server that accepts connections but can't answer counts as down.
Response time
How long connect and query took together, with anomaly detection for this hour of the week.
Dependency folding
Point your API's monitor at the database as its parent, and one outage becomes one alert instead of twelve.
And the alert actually arrives
Most monitoring tools are fine until the one time the alert doesn't arrive. Uplance is built so that can't happen quietly.
- Every alert is written to an outbox before it is sent, and retried for about fifty minutes. A restart doesn't lose it.
- Once a week, a real test event is sent through every channel you've configured, and you're told which ones failed.
- If the alert pipeline is in trouble, the dashboard says so — it isn't buried in a log file.
- If the checker itself stops, the health endpoint returns 503 so an outside heartbeat can page you, and your status page says monitoring has stopped instead of showing stale green.
Common questions
- What privileges does the monitoring user need?
- Only the right to connect. Create a user with no grants beyond that; the check proves the server answers and never touches your tables.
- Does it work with MariaDB?
- Yes, and with the managed MySQL services. Give it a connection string and it connects the way any client would.
- Can it reach a database on a private network?
- Not from the hosted checkers — they refuse private, loopback and cloud-metadata addresses on purpose. A remote agent inside your network runs the check there and reports only the result out.
- Where are the credentials kept?
- Encrypted with AES-256-GCM at rest, decrypted only to run the check, and never written to logs, check history or alert payloads.
Also monitored
Know it's down before your customers do.
10 monitors, 30 days of history, no credit card