Analysis of a Percona XtraDB Cluster (PXC) Flow‑Control Incident Triggered by Disk I/O and Large TEXT Fields
This article examines a real‑world Percona XtraDB Cluster (PXC) fault where simultaneous write‑node connection spikes were caused by flow‑control triggered by 100% disk I/O on a shared read node, ultimately traced to slow sorting queries involving a mediumtext column, and offers diagnostic steps and mitigation strategies.