Observability
Observability refers to the extent to which the internal state and behavior of a system can be understood, monitored, and analyzed from the outside, typically by developers and DevOps. It focuses on providing insight into how a system is performing, what is happening inside it, and how it is interacting with its environment.
The goal of observability is to make it easier to diagnose and resolve issues, optimize performance, and gain insights into the system's behavior. It is especially important in modern, complex, and distributed systems, where understanding the interactions between different services and components can be challenging. DevOps Research and Assessment (DORA) research shows that a comprehensive monitoring and observability solution, along with several other technical practices, positively contributes to the management of software infrastructure.
YugabyteDB provides several components and features that you can use to actively monitor your system and diagnose issues quickly.
Metrics
Use metrics to track trends and identify performance issues, and manage the system's performance and reliability.
YugabyteDB exports various metrics, which are effectively quantitative measurements of the cluster's performance and behavior. These metrics include details on latency, connections, cache, consensus, replication, response times, resource usage, and more:
- Throughput and latency metrics
- Connection metrics
- Cache and storage subsystem metrics
- Raft and distributed system metrics
- Replication metrics
- YB-Master metrics
Alerting and monitoring
Monitoring involves continuously checking the system's health and performance and notifying stakeholders if any issues arise. For this, you can set up automated alerts based on predefined thresholds or conditions. All metrics exposed by YugabyteDB are exportable to third-party monitoring tools like Prometheus and Grafana which provide industry-standard alerting functionalities.
Visualization and analysis
YugabyteDB provides dashboards that include charts, graphs, and other visual representations of the system's state and performance. yugabyted starts a web-UI on port 15433 that displays different charts for various metrics.
You can also export the metrics provided by YugabyteDB onto third-party visualization tools like Prometheus and Grafana as per the needs of your organization.
Use cases
Operational monitoring
You can build an application health dashboard for your critical applications using key operational signals that are constantly monitored. Add alerts for DevOps or SRE teams so they can act quickly in case of an event to ensure business continuity. The application health dashboard collects signals, metrics from YugabyteDB, and other systems that power your application, such as APIs, web app, SDK, and so on.
Performance troubleshooting
Database administrators and application developers need to be able to troubleshoot issues, perform root cause analysis, and issue fixes. You can create a dashboard to monitor an observed issue causing temporal, gradual, or systemic performance degradation, or application failure. To conduct root cause analysis, issue-dependent deep observability metrics in a specific area are typically used. These metrics are consumed at the time of root cause analysis and operating teams fall back to a health dashboard after the issue is identified, the fix is monitored, and the issue is resolved.
Object monitoring
Monitor specific parts of application behavior continuously after a new feature launch, during maintenance windows, during application upgrades, and more. The metrics can be system-wide or specific to the object of interest, such as a YugabyteDB cluster, node, tablet, geography, users, tenant, and more.
Logging
Logs from different services, such as the YB-TServer and YB-Master provide a historical record of what has happened and can be very helpful in debugging and troubleshooting. These logs are rotated regularly, based on their size as configured. See Logs management.
Query-level metrics
The following table describes views in YSQL you can use to monitor and tune query performance.
| View | Description |
|---|---|
| pg_stat_statements | Get query statistics (such as the time spent by a query) |
| pg_stat_activity | View and analyze live queries |
| yb_local_tablets | Get YSQL/YCQL and tablet metadata details |
| yb_terminated_queries | Identify terminated queries |
| pg_stat_progress_copy | Get the status of a COPY command execution |
| pg_locks | Get information on locks held by a transaction |
To get more details about the various steps of a query execution, use the Explain Analyze command.
Active Session History
Active Session History (ASH) offers insight into current and past system activity by periodically sampling session behavior in the database. ASH functionality extends to YSQL, YCQL, and YB-TServer processes, and helps you to conduct analytical queries, perform aggregations, and troubleshoot performance issues.