- Print
- PDF
Meaning
This alert is triggered when call conversations are successfully analyzed but are not being indexed at the same rate. This suggests that the indexing process is falling behind, causing a growing backlog between analyzed and indexed data. If the backlog increases continuously for 10 minutes, a warning-level alert is raised.
Full context
Knovvu Analytics tracks the progression of conversations from analysis to indexing. When the system observes that analyzed conversations accumulate without being indexed promptly, this could indicate indexing delays, processing bottlenecks, or component failures.
The alert compares the indexing lag over two consecutive one-hour time windows. If the difference is growing, the system raises this alert.
Impact
If indexing cannot keep up with analyzed conversations:
- Conversations will not be searchable or visible in dashboards.
- Downstream systems that depend on indexed data (e.g., search, reports, insights) may operate on stale or incomplete data.
- The backlog may grow to the point of exhausting queue capacity or delaying time-sensitive analytics.
- Operational visibility into agent-customer interactions may be lost temporarily.
Diagnosis
- Review dashboards to compare the rate of analyzed vs. indexed call conversations over the past few hours.
- Check the health and performance of the indexing management service (
ca-index-manager
) to verify jobs are being created and dispatched correctly. - Focus on the
ca-indexer
component, which performs the actual heavy-lifting of indexing. Investigate resource usage (CPU, memory, disk I/O), thread pools, and indexing throughput. - Inspect logs from both components for indexing errors, throttling, or task failures.
- Determine whether any recent deployments or infrastructure changes have impacted performance.
Mitigation
- Scale out the
ca-indexer
service if it is under heavy load and unable to keep up. This component is typically designed for horizontal scalability. - Restart or tune the
ca-indexer
to optimize performance under current traffic patterns. - Ensure that
ca-index-manager
is properly dispatching indexing jobs and not blocked or misconfigured. - Investigate potential data issues (e.g., oversized payloads, malformed fields) that could slow down the indexing process.
- Monitor the backlog size and verify it starts to shrink after action is taken.
- Escalate to engineering if indexing delays persist or backlog trends continue rising.