Understanding the "DBSS - SQL I/O Errors" Rule in Foglight for SQL Server (4379526)

The Foglight rule “DBSS - SQL I/O Errors” raises an alarm when SQL Server reports one or more I/O errors. Customers may observe alarms with messages such as:

"SQL I/O Errors. X I/O errors have been encountered. Please check SQL Server log."

This article explains what this rule monitors, whether it detects specific SQL Server or Windows Event IDs (e.g., Event ID 833), and how the threshold works.

This rule is triggered when SQL Server reports disk I/O errors during Foglight’s data collection cycle. These errors generally indicate that SQL Server was unable to read from or write to disk due to a hardware or system-level failure. Examples include:

• Failed write attempts to database or log files

• OS-level disk or controller issues

• Hardware faults that interrupt disk I/O

The rule does not monitor SQL Server performance warnings such as:

Event ID 833: SQL Server has encountered I/O requests taking longer than 15 seconds to complete...

The rule also does not monitor specific Windows or SQL Server Event IDs directly. Instead, it uses internal SQL Server error counters to detect I/O failures.

What the Rule Monitors

• The rule uses the metric #sql_io_errors#, which tracks the number of new SQL Server I/O errors detected since the previous collection. This metric comes from SQL Server as the result of SELECT @@TOTAL_ERRORS; as described in the Microsoft documentation here.
• These errors are reported by SQL Server internally and typically relate to:
  • Operating system errors like “I/O device error”
  • SQL Server errors such as 823, 824, or 825
  • Write or read failures to data or log files

These SQL Server errors may generate related entries in the Windows Event Log (e.g., Event ID 17053 for write failures), but the rule does not detect or filter by specific Event IDs.

What the Rule Does Not Monitor

• Performance-based warnings, such as:
  • Event ID 833: Indicates slow I/O, not failure
  • Disk latency or stall conditions
• Any specific Event ID or Windows log entry

Threshold Unit of Measure

• The threshold is measured in whole-number counts of new I/O errors.
• Fractional values (e.g., 0.5) are invalid and may lead to unpredictable behavior.
• Use a threshold of 1 to trigger an alarm on any I/O error.

To Monitor Event IDs Like 833

To monitor specific SQL Server or Windows Event IDs, consider:

1. Using Foglight’s Log Monitor or Windows Event Log Integration: Create custom rules that trigger based on event ID or log message content.
2. Creating a Custom Rule: Implement a custom Foglight script to analyze Windows logs or SQL logs for targeted events.

• SQL Server I/O errors are serious and should be investigated promptly to prevent data loss or corruption.
• To monitor slow I/O, consider using:
  • sys.dm_io_virtual_file_stats for average I/O stall times
  • Windows Performance Monitor (PerfMon)
  • Extended Events or system monitoring tools