Class FcNodeStats

java.lang.Object
org.frontcache.core.FcNodeStats
public class FcNodeStats extends Object
A cheap snapshot of the JVM this node runs in: heap, process CPU, uptime. The companion of CacheStats - that one answers "what is in the cache", this one answers "how is the process holding up" - and it follows the same two rules, for the same reason. Everything is read from beans the JVM already maintains, because this is polled by a console that never stops polling; and every field can be UNKNOWN, because a number this JVM cannot produce has to be distinguishable from a real zero. A heap gauge flatlining at 0 is a lie somebody eventually builds an alert on.

Micrometer-free on purpose

Micrometer ships JVM binders that produce all of this. It is deliberately not used here: metrics export is optional and off by default (front-cache.metrics.export=none), and the management API has to answer whether or not anyone enabled it. Same reasoning as CacheStats, which exposes plain numbers so that org.frontcache.metrics stays the only package that knows what a meter is.

Nothing here throws

Every reader is guarded. This feeds an endpoint on a poll loop, and a management response that fails because one bean was unavailable is worse than one reporting UNKNOWN for that field.

  • Field Details

    • UNKNOWN

      public static final long UNKNOWN
      The JVM does not report this number. Distinct from zero.
      See Also:

    • UNKNOWN_LOAD

      public static final double UNKNOWN_LOAD
      UNKNOWN for the CPU load, which is a fraction rather than a count.
      See Also:

    • NOT_TRACKED

      public static final FcNodeStats NOT_TRACKED

  • Constructor Details

    • FcNodeStats

      public FcNodeStats(long heapUsedBytes, long heapMaxBytes, double cpuProcessLoad, long uptimeMillis)

  • Method Details

    • getHeapUsedBytes

      public long getHeapUsedBytes()
      Heap currently in use, or UNKNOWN.

    • getHeapMaxBytes

      public long getHeapMaxBytes()
      The heap ceiling, or UNKNOWN. MemoryUsage.getMax() already returns -1 when the maximum is undefined, which is the same value UNKNOWN carries - so an unbounded heap reports as untracked rather than as a ceiling of zero, and a caller rendering "used / max" has nothing to divide by. That is correct: there is no percentage to show.

    • getCpuProcessLoad

      public double getCpuProcessLoad()
      Recent CPU used by THIS JVM as a fraction of all available cores (0..1), or UNKNOWN_LOAD. Process CPU, not system CPU: on a shared host the system figure answers a question about the neighbours, not about this node. The HotSpot bean computes this over its own internal window and returns a negative value until it has two samples, so the first reading after a restart is UNKNOWN_LOAD. That is why the number is taken from the bean rather than from a getProcessCpuTime() delta kept here: a delta would be "since whoever last asked", which quietly means something different the moment a second console starts polling.

    • getUptimeMillis

      public long getUptimeMillis()
      Milliseconds since JVM start, or UNKNOWN.

    • isTracked

      public boolean isTracked(long value)

    • isTracked

      public boolean isTracked(double load)

    • current

      public static FcNodeStats current()
      Reads the JVM now. Never throws, never blocks.

    • toString

      public String toString()
      Overrides:
      toString in class Object