Root cause correlation
Root Cause Correlation (RCC) finds a root cause by automatically correlating metrics, logs, and event information for supported symptoms on production instances for the last 24 hours.
RCC symptom categories
The RCC feature is available for self-service alerts and Critical or Warning performance symptom categories:
- Memory
- Longest running sessions
- Slow transactions
- Cache flush
- Database locks
- Database impacts
The table describes the symptom categories and the corresponding alerts that the RCC engine detects.
| Symptoms categories | Description | Corresponding alert |
|---|---|---|
| Database Impact | Helps you to identify and address extended SQL queries impacting database performance, tied to high execution time, or increased volumes. The query patterns give snapshots of 30-minute and 60-minute durations from the time that the impact is observed on query execution times. | Database response time |
| Cache Flush | Cache flushes and node restarts are detected, as well as high service saturation levels that may have occurred around the time a performance alert was triggered. | Default semaphore mean |
| Longest Running Session | - Searches for the top long-running sessions in mean time to recover \(MTTR\) logs - Identifies the top transaction pattern hash with the highest processing times and then the transaction IDs. | Default semaphore mean |
| Slow Transactions | - Identifies the top long-running transactions using the total duration, including ACL time, SQL time, CPU time, processing time, BR time, and script time. - Returns the transaction IDs, the pattern hash, and these metrics to help you identify the specific causes of long-running transactions. | Default semaphore mean |
| Memory |
Note: Users are advised to review these long-running or frequently recurring threads. | Node Garbage Collection Time |
| DB locks | The RCC engine monitors innodb\_row\_lock\_waits and threads\_running to detect anomalous database lock events that occur when a database operation requires exclusive access. | Threads running |
Note: As soon as any of the aforementioned alerts or Critical or Warning performance are identified, the system automatically generates an RCA report after waiting for a maximum of 10 minutes, depending on the conditions.
Parent Topic:Monitor instance performance with IO