Writing the same key many times can cause a single KV range to grow beyond the configured max range size and still be unsplittable. Large ranges often have various side effects that are hard to predict. When diagnosing instability within a cluster, it would helpful to surface all "large" ranges. This came up in cockroachlabs/support#2859 where the large range had several effects including high CPU on nodes with replicas of the large range, and an overloaded LSM on a node that received a snapshot of the large range.
One idea is that we could surface large ranges in the 'hot ranges' page/dump.
Writing the same key many times can cause a single KV range to grow beyond the configured max range size and still be unsplittable. Large ranges often have various side effects that are hard to predict. When diagnosing instability within a cluster, it would helpful to surface all "large" ranges. This came up in cockroachlabs/support#2859 where the large range had several effects including high CPU on nodes with replicas of the large range, and an overloaded LSM on a node that received a snapshot of the large range.
One idea is that we could surface large ranges in the 'hot ranges' page/dump.
Jira issue: CRDB-36655
Epic CRDB-32398