Server-Generated Alerts
Published by : Obay Salah , November 19, 2024
It is a system that monitors the database, the Instance, and the User Sessions, issuing warnings at certain thresholds and when specific events occur.
These warnings result from a large number of metrics that are continuously calculated and monitored, and the monitoring process is done by the MMON (Manageability Monitor).
Metric: Refers to statistics that are converted into meaningful numbers. In previous releases of databases, the process of collecting statistics and converting them into useful metrics was usually done by the database administrator. Now, Oracle does this and stores the results in the Automatic Workload Repository (AWR) in the SYSAUX Tablespace.
However, it must be noted that all metrics are not calculated repeatedly the same way; rather, they are calculated with different frequencies. For example, the Tablespace Space Metric is calculated every ten minutes because space issues do not appear suddenly, while CPU usage is monitored continuously every second.
Data Metric writes to the AWR every hour, and generally, there are more than 120 metrics.
Generally, Alerts can be divided into Threshold and Nonthreshold. For example, the Usage Tablespace uses a Threshold Alert. Typically, you can configure the Alert System to send a Warning Alert when the Tablespace reaches 95% full, and to send a Critical Alert when it reaches 99% full. Based on these alerts, the database administrator can take appropriate actions.
Nonthreshold refers to warnings that notify you of unexpected events, such as "ORA-1555: snapshot too old," which indicates a failure in querying data in the Undo Data because this data is outdated and has been overwritten. When this error occurs, a quick solution is required. This type of error is also called (stateless) because it does not persist but rather occurs and then goes away.
Additionally, the database administrator can inquire about metrics via V$ALERT_TYPES.
There are more than 120 metrics in the database, and all of these metrics can be configured by the database administrator for Warning and Critical Alerts.
The Instance can collect statistics itself through the Alert System, provided that the STATISTICS_LEVEL variable is set to TYPICAL or ALL; if this variable is set to BASIC, many statistics will fail to work.
During the statistics and metrics process, the MMON compares these metrics with the configured Threshold Alert. If the metrics exceed the thresholds, the MMON writes a message to the queue, and if a Nonthreshold event occurs, the MMON also writes a message to the queue. This queue, a feature in the database, allows a group of Sessions to communicate with each other and is called ALERT_QUEUE.
Here, a group of Sessions can participate in this queue, which is cleaned up after receiving all these messages.
The database administrator can inquire about the Alert via DBA_OUTSTANDING_ALERTS, which will show all alerts that have occurred and have not been dealt with yet. However, after that, they are stored in the DBA_ALERT_HISTORY for reviewing all alerts.
Comments
no comment yet!