Health Check of a Medulla server
This procedure allows you to quickly verify that a Medulla server and its main components are functioning properly.
In particular, it can be used:
- after a server reboot;
- after a maintenance operation;
- after a Medulla update;
- after a configuration change;
- when a general platform malfunction is suspected.
Note:
The purpose of this Health Check is to quickly confirm that the main Medulla components are operational. If any of the checks fail, review the logs for the affected component before restarting or modifying the service.
1. Verify access to the Medulla interface
Start by logging in to the Medulla web interface.
In particular, verify that:
- the login page is accessible;
- authentication is working;
- the main interface loads correctly;
- the computers appear in the inventory.
Expected result:
The administrator can log in to Medulla and view the computer inventory as expected.
2. Check the status of the main Medulla services
Check the status of the main Medulla services using systemctl.
systemctl status ejabberd
systemctl status mmc-agent
systemctl status pulse-xmpp-agent-relay
systemctl status pulse2-register-pxe
systemctl status pulse2-package-server
systemctl status pulse-xmpp-master-substitute-master
systemctl status pulse-xmpp-master-substitute-logger
systemctl status pulse-xmpp-master-substitute-updates
systemctl status pulse-xmpp-master-substitute-assessor
systemctl status pulse-xmpp-master-substitute-inventory
systemctl status pulse-xmpp-master-substitute-deployment
systemctl status pulse-xmpp-master-substitute-monitoring
systemctl status pulse-xmpp-master-substitute-registration
systemctl status pulse-xmpp-master-substitute-subscription
For a quicker check, use:
systemctl is-active SERVICE_NAME
The expected result is:
active
Caution:
Depending on the Medulla installation and the enabled features, some services may not be present. The absence of an optional service does not necessarily indicate a malfunction.
3. Quickly identify services with errors
To find services that are currently failing:
systemctl --failed
If a Medulla service appears in this list, check its status:
systemctl status SERVICE_NAME
then its latest logs:
journalctl -u SERVICE_NAME -n 100
Tip:
Do not automatically restart a service that is failing before checking its logs. The messages present before the restart may be essential for identifying the cause of the incident.
4. Check ejabberd and XMPP
Medulla uses XMPP to communicate with its agents and several of its internal components.
Verify that the ejabberd service is running:
systemctl is-active ejabberd
The expected result is:
active
Next, check the number of users currently connected:
/usr/bin/ejabberdctl stats onlineusers
and the number of registered users:
/usr/bin/ejabberdctl stats registeredusers
To view the currently connected users:
ejabberdctl connected_users
Expected result:
ejabberd is active, and the expected Medulla agents and components are connected to XMPP.
5. Check MariaDB
Verify that the MariaDB service is running:
systemctl is-active mariadb
The expected result is:
active
Next, verify that MariaDB is responding correctly to queries:
mariadb -e "SELECT 1;"
The expected result includes:
1
1
6. Check the package server
Check the package server service:
systemctl is-active pulse2-package-server
The expected result is:
active
Also verify from the Medulla interface that an existing package is correctly visible.
7. Check the relay server
If the server also acts as a relay, check:
systemctl is-active pulse-xmpp-agent-relay
The expected result is:
active
In an architecture with multiple relays, also verify that the expected relays are accessible.
If a relay is unavailable, this can cause, among other things:
ABORT RELAY DOWN
ABORT ALTERNATIVE RELAYS DOWN
8. Check Syncthing
When Syncthing is used for package synchronization, check its status:
systemctl is-active syncthing@syncthing
The expected result is:
active
Caution:
An active Syncthing service alone does not guarantee that all packages are properly synchronized. If there is a transfer error or a missing package on a relay, also verify that synchronization has actually taken place.
9. Check disk space
Check the available space on the file systems:
df -h
Pay particular attention to the file system containing the deployment packages.
Important:
A full repository on a relay may prevent the synchronization or distribution of new packages without necessarily causing a service to stop.
10. Check server resources
Check available memory:
free -h
Next, check the server load:
uptime
For a real-time view:
top
Check the following in particular:
- available memory;
- swap usage;
- system load;
- whether there are any processes consuming an abnormal amount of resources.
11. Verify that a computer is communicating with Medulla
From the Medulla interface, select a computer known to be powered on and verify that it appears online.
This check provides a simple way to validate the communication chain:
Agent
↓
Réseau
↓
XMPP / ejabberd
↓
Medulla
If all computers appear offline even though they are powered on, first check for a system-wide communication or service issue.
12. Check the inventory
Select a test computer and verify that its inventory information is available in Medulla.
Check, for example:
- the hostname;
- the operating system;
- network information;
- installed software.
Expected result:
The computer's inventory information is available and consistent.
13. Test a deployment
After major maintenance or an update, it is recommended to perform a deployment on a test computer.
Preferably use a simple package that does not have a significant impact on functionality.
The deployment must result in:
DEPLOYMENT SUCCESS
This test validates, in particular:
- communication with the agent;
- relay assignment;
- package availability;
- the transfer;
- execution on the target computer.
14. Check for recent errors
Even when all services are active, check the logs for recent errors.
For a service:
journalctl -u SERVICE_NAME -n 100
To view errors since the last startup:
journalctl -p err -b
Note:
A service may be active while still experiencing functional errors. Therefore, the systemd status is not always sufficient to confirm that the component is functioning properly.
15. Quick checklist
| Check | Expected result |
|---|---|
| Web interface | Accessible |
| Authentication | Functional |
| Medulla services | Active |
systemctl --failed |
No critical Medulla services are down |
| ejabberd | Active |
| XMPP | Components expected to be connected |
| MariaDB | Active and accessible |
| Package server | Active |
| Relay server | Active and accessible |
| Syncthing | Active if in use |
| Disk space | Sufficient available space |
| Memory / load | No overload |
| Test computer | Online |
| Inventory | Available |
| Test deployment | DEPLOYMENT SUCCESS |
16. What should I do if a check fails?
| Problem detected | Further troubleshooting |
|---|---|
| Medulla service is down | Check systemctl status, then journalctl |
| ejabberd is unavailable | Diagnose the XMPP service |
Component missing from connected_users |
Diagnose its XMPP connection |
| MariaDB is unavailable | View MariaDB status and logs |
| Relay unavailable | Diagnose the relay server |
| Package missing from a relay | Check Syncthing and synchronization |
| Computer offline | See the FAQ “Why does a computer appear offline?” |
| Computer not registered | See the FAQ “Why won’t a computer register?” |
| Deployment error | Check the deployment status and audit |
17. Summary
A Medulla server can be considered generally operational when:
- the web interface is accessible;
- the main services are active;
- ejabberd is running;
- the expected components are connected to XMPP;
- MariaDB is running;
- the relay servers are available;
- the packages are accessible and synchronized;
- system resources are not overloaded;
- at least one computer is communicating normally with Medulla;
- the inventory is working;
- a test deployment can be performed successfully.
Recommendation:
After maintenance or a major update, don’t just check that the services are active. Also verify end-to-end functionality with an online computer, an inventory update, and a test deployment.