OVNPM 1.5.2 - Performance Agent for NonStop Server Monitoring Guide

Configuring Loop Detection 57
5 Configuring Loop Detection
5-1 Introduction
This feature detects and reports any process using excessive CPU or memory resources, conditions that typically
indicate a runaway program loop. Runaway Loop Detection operates as part of the SystemAlert module.
Runaway memory loops and CPU loops are two of the most severe problems that may occur during the life of a
process. Ordinarily, these conditions are very difficult to detect. OVNPM can detect these situations and
immediately reports them to avoid dramatic system degradation.
OVNPM collects data at regular and frequent intervals (either every 15 or 30 seconds) and checks the state of
every process.
Detection is based on changes in key metrics in a process that may have occurred during the collection interval
compared to one or more threshold values. If these thresholds are reached, OVNPM considers the process to be
in one of the runaway loop states (CPU or memory) and SystemAlert issues an alarm.
OVNPM provides default threshold values that can be customized to suit the needs of an installation. The DCPRM
file stores threshold values.
5-2 CPU Loop Detection
NOTE:
The HP NonStop Guardian system reduces the priority of processes that are consuming excessive CPU
resources. Priority 1 is the lowest priority status.
A process is detected as being in a runaway Cpu-loop state if the following conditions are met:
• Its priority is equal to 1 OR
its priority has decreased since it was last checked
AND
• Its Cpu-busy % exceeds the Cpu-loop threshold parameter (CpuLoopThreshold) set in the DCPRM file
Examples
These examples illustrate processes that are detected as runaway CPU loops:
Priority = 1 AND Cpu-Busy = 75% (CpuLoopThreshold = 70)
Priority = 70 (priority in previous collection interval = 72) AND Cpu-Busy = 72% (CpuLoopThreshold = 70)
5-3 Memory Loop Detection
A process is detected as being in a runaway memory-loop (Mem-Loop) state if the following conditions are met:
• It acquires more incremental memory during a collection interval than is set in the memory increment
threshold (MemIncStep parameter)
AND
• It continues to exceed the MemIncStep threshold for at least the number of consecutive collection intervals
set in the memory-interval threshold parameter (MemIncNb)
Examples
These examples illustrate processes that are detected as runaway memory loops:
MemIncStep threshold = 6 and MemIncNb threshold = 2
• 13:30:30 a process uses 3450 memory pages
• 13:31:00 this same process uses 3500 memory pages
• (3500-3450) is greater than 6 (MemIncStep)
• 13:31:30 this same process uses 3700 memory pages
• (3700-3500) is greater than 6 (MemIncStep)
• 13:32:00 this same process uses 3750 memory pages
• (3750-3700) is greater than 6 (MemIncStep)
Configuring Loop Detection