Understanding Data Gaps in Graphs

When monitoring your servers with Server Scout, you may occasionally notice gaps in your monitoring graphs. Understanding why these gaps appear and how they're handled is crucial for accurate system monitoring and troubleshooting.

How Gap Detection Works

Server Scout employs an intelligent gap detection system that identifies when monitoring data is missing. The system flags a gap when no data point exists for more than three times the expected collection interval.

For example, if your agent is configured to collect metrics every 5 seconds, a gap will appear in the graph after 15 seconds of missing data. Similarly, for 30-second intervals, gaps become visible after 90 seconds of silence.

This approach ensures that brief network hiccups or momentary delays don't create unnecessary visual disruptions in your graphs, whilst still clearly highlighting genuine monitoring interruptions.

Common Causes of Data Gaps

Several scenarios can lead to gaps in your monitoring data:

Agent Restart or Shutdown When the Server Scout agent is restarted or stopped for maintenance, data collection ceases temporarily. This creates the most common type of gap you'll encounter.

Network Connectivity Issues If your server loses network connectivity or experiences routing problems, the agent cannot transmit collected metrics to the Server Scout platform, resulting in gaps.

Server Reboots System reboots naturally cause monitoring gaps as the agent stops during shutdown and takes time to restart after boot.

API Downtime Whilst rare, temporary issues with Server Scout's data ingestion API can prevent metric submission, though the agent's spooling mechanism helps minimise this impact.

Data Retention and Time Buckets

Server Scout uses a tiered data retention system that affects how gaps appear at different time scales:

1. Raw data is retained for 24 hours at full resolution
2. 1-minute aggregated data is kept for 7 days
3. Older data is progressively pruned to manage storage efficiently

When viewing graphs over different time periods, the system uses appropriate time buckets for downsampling. For instance:

• Last 6 hours: Raw data points every 5-30 seconds
• Last 24 hours: 1-minute buckets
• Last 7 days: 5-minute buckets
• Longer periods: 15-minute or hourly buckets

This downsampling can affect gap appearance. A 2-minute outage might be clearly visible when viewing the last hour but could be less apparent when examining a week-long timeline with 15-minute buckets.

How Data Spooling Minimises Gaps

The Server Scout agent includes a sophisticated data spooling mechanism designed to minimise gaps during network interruptions. When connectivity is lost, the agent continues collecting metrics locally and stores them in a temporary buffer.

Once connectivity is restored, the agent automatically transmits the buffered data to fill in the gaps. This means that brief network outages often won't result in permanent data loss, though you may see temporary gaps that resolve once the backlogged data is processed.

The spooling system has configurable limits to prevent excessive disk usage during extended outages. If the buffer reaches capacity, older spooled data may be discarded to make room for current metrics.

Interpreting Gaps in Context

When analysing gaps in your monitoring data, consider the context:

• Short gaps (under 5 minutes) are typically network-related and may self-resolve through spooling
• Medium gaps (5-30 minutes) often indicate agent restarts or brief system issues
• Long gaps (over 30 minutes) usually suggest significant problems like server downtime or extended maintenance

Best Practices for Gap Management

To minimise monitoring gaps:

1. Configure your agent to start automatically on system boot
2. Monitor your network connectivity and address routing issues promptly
3. Plan maintenance windows and expect gaps during scheduled downtime
4. Review gap patterns to identify recurring issues that need attention

Understanding these gap behaviours helps you distinguish between genuine system problems and normal monitoring interruptions, leading to more effective server management and troubleshooting.