Package org.exoplatform.addons.documents.cleanup.util

Class CleanupEtaUtil

java.lang.Object

org.exoplatform.addons.documents.cleanup.util.CleanupEtaUtil

public class CleanupEtaUtil
extends Object

Single definition of the throughput ETA of a cleanup worker, shared by the dry-run scan and the execution purge — the two workers used to carry the very same arithmetic verbatim, so a change (or a regression) in one of them silently diverged from the other.

The estimate is deliberately based on the throughput observed SINCE THIS RUN STARTED, not since the campaign started: a resumed worker must not inherit the throughput of the run that was interrupted (a machine that was slow yesterday says nothing about today's remaining time). It is CUMULATIVE over that run and not windowed — it was once described here as 'rolling', which it never was.

UNIT-AGNOSTIC on purpose: the arithmetic is 'work done over time elapsed', and the caller chooses what a unit of work is. The scan counts NODES, whose cost is near-uniform. The purge counts BYTES (CleanupExecutionService#purgeEtaSeconds), because its items differ in cost by orders of magnitude and a count-based average predicts the items already met rather than the ones left.

Method Summary

Modifier and Type	Method	Description
static long	computeEtaSeconds(long startTime, long processedAtStart, long processed, long total)	Same as computeEtaSeconds(long, long, long, long, long) evaluated against the current server clock.
static long	computeEtaSeconds(long startTime, long processedAtStart, long processed, long total, long nowMillis)	Remaining time of a run, from the throughput measured since it started.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

computeEtaSeconds

public static long computeEtaSeconds(long startTime,
 long processedAtStart,
 long processed,
 long total)

Same as computeEtaSeconds(long, long, long, long, long) evaluated against the current server clock.

Parameters:

startTime - epoch millis at which THIS run started

processedAtStart - work already done when this run started (a resumed worker starts above zero)

processed - work done so far, including processedAtStart

total - denominator of the run, in the SAME unit

Returns:

estimated remaining seconds, 0 when it cannot be estimated yet

computeEtaSeconds

public static long computeEtaSeconds(long startTime,
 long processedAtStart,
 long processed,
 long total,
 long nowMillis)

Remaining time of a run, from the throughput measured since it started. Short-circuits to 0 — 'unknown', never a negative or infinite value — whenever the throughput isn't measurable yet: no time elapsed, or nothing processed since the run started (the very first batch of a resumed worker). A numerator that already reached (or passed) the denominator also yields 0.

Parameters:

startTime - epoch millis at which THIS run started

processedAtStart - items already processed when this run started

processed - items processed so far, including processedAtStart

total - denominator of the run

nowMillis - evaluation time (epoch millis)

Returns:

estimated remaining seconds, 0 when it cannot be estimated yet