From: Priya Nair <priya.nair@example.com>
To: Storage Engineering <storage-eng@example.com>
Subject: Read latency regression after replica rebalancing
Date: Fri, 31 Jul 2026 10:42:18 -0400

Hi team,

We saw p99 read latency increase from 38 ms to 121 ms in the us-east cluster shortly after yesterday’s replica rebalancing. Error rates stayed flat, but requests that crossed availability zones were disproportionately affected.

The current theory is that several hot partitions ended up with leaders in us-east-1c while most clients are running in us-east-1a. I’ve temporarily moved the three busiest leaders and latency is already trending down.

Could someone review the placement policy before the next rebalance? We should probably account for request locality as well as disk utilization when choosing leaders.

Thanks,
Priya