Reduce log volume in evacuation check methods

[kabragaurav]

Problem

EKG log volume check is failing for helix.ambry and helix.espresso2 during canary deployments in EI cluster:

"Total log volume should not change by more than 100% when experiment host produces log volume more than 1 MB/min"

Observation: The rule is flaky (failing mostly but sometimes passing). Started failing from version 11.3.0.31 (Dec 12).

Jira: CICP-3460

Root Cause Analysis

After analyzing logging changes in past ~45 days, identified culprit: Commit 99b5c158e introduced 6 new INFO logs in ZKHelixAdmin.java's instanceHasCurrentStateOrMessage() method.

The problem: This method is called on every REST API poll to isEvacuateFinished(), isInstanceDrained(), and isReadyForPreparingJoiningCluster(). The partition-counting log was firing on every successful call, even when evacuation was complete.

Why only Ambry & Espresso2 in EI?

• Storage systems with frequent node maintenance (evacuations/swaps)
• Many instances being monitored simultaneously
• High polling frequency from ACM automation

Solution

Made partition-counting logs conditional - only fire when partitions are actually blocking evacuation:

if (hasRemainingPartitions) {
    logger.info("Instance {} in cluster {} has {} partitions after applying {} exclusions...");
}