Concept
Operations Alerts
The warnings that go to system administrators and the alert addresses after unusual unpublishing, when an account is locked and when a task, a harvest or a background job fails.
Some events can point to a fault or an attack. When they happen, JournTech sends a panel notification to all active system administrators. The notification appears under the bell in the top bar and on the Notifications page. The same warning goes by e-mail to the addresses in the Alert e-mail addresses field. If Telegram is set up for the installation, the warning is sent to Telegram too. The warning is also written to the application log.
Events that create an alert
| Event | Frequency limit |
|---|---|
| Many catalogue records or journal issues unpublished, records retracted or previously published records deleted within a short time | One alert per window |
| An account locked after repeated failed sign-ins | At most one alert per account per hour |
| Many accounts locked within one hour | After 10 account alerts, one combined alert for that hour |
| A periodic background task failing | At most one alert per task per hour |
| The ISBN range message not refreshed for more than three weeks | One alert a day until it is refreshed |
| A harvest from DergiPark (Türkiye's national journal platform) failing | At most one alert per journal per hour |
| A thesis harvest run failing | At most one alert per hour |
| A removal check started by the system not finding many theses on the repository's list | At most one alert per hour |
| Background jobs failing after all their attempts, or 50 jobs waiting for more than 30 minutes | At most one alert per hour |
| A test alert sent from the settings | At most one alert per minute |
The alert about failed background jobs counts every job that failed since the previous such alert. Jobs that fail before the hour is up therefore appear in the next alert.
Alert e-mail addresses
The addresses are on the Operations alerts card on the Site settings → Workflow tab. Write one address per line and press Save. Up to 10 addresses can be given. The IT department's shared address can go here. System administrators whose own addresses are not on this list see the alerts in the panel only.
Alert e-mails are written in the installation's default language and sent with the SMTP settings of the E-mail tab. While e-mail delivery is off, they are not sent and wait on the System health → E-mail outbox tab. Alert e-mails that were sent appear on this tab too.
Test alert
After saving the addresses, press Send a test alert below the address field. The test alert takes the same path as a real alert. The system administrators get a panel notification and the saved addresses get an e-mail. The button does not work while there is an unsaved change. At most one test alert can be sent per minute. Every test is written to the audit log.
Setting the unpublishing alert
The threshold and the window are in the Mass take-down alert part of the same card. More actions than the threshold within the window create an alert. The default threshold is 5 actions, and the default window is 10 minutes. Entering 0 as the threshold turns the alert off. Deleting a record that was never published does not count.
Alert content
Alert texts carry only IDs, counts, task names and journal names. They contain no e-mail addresses, passwords, codes or keys.
The link in an account lock alert opens the account's user page. If the password was right and the two-factor authentication codes were wrong, the alert says so separately. The locked-out person is told by e-mail too. If the attempts did not come from the person, look into the account and disable it if needed. The steps are in Managing User Accounts.
The link in an unpublishing alert opens the audit log. If the actions were not expected, check the account that carried them out. The audit log is explained in Audit Log.
Alert templates and records
The notification and e-mail texts are the templates on the E-mail templates screen whose names start with
"Operations alert" and "Security alert". If a template's e-mail is switched off, that alert does not go to the
addresses, but the panel notification still goes out. Every alert that is sent has an ops_alert job in the
System health → Background jobs list.
Monitoring systems
The institution's monitoring system (Prometheus, Zabbix or similar) can read JournTech's health, error rate and response
time measurements at the /metrics address. This address opens when an access key is set during installation, and it
answers only requests that send this key. The team that runs the installation sets the key. The server's disk use and
the certificate's validity are measured by the monitoring system's own tools.