# Stale-Read Latency Spikes Due to Heavy Schema Loading

Source note: adapted from the Feishu `Storage oncall handbook` <https://pingcap.feishu.cn/wiki/RNzIwCexFiA1uXkRCvzcsIocnVb> and normalized into the transaction handbook.

- Phenomenon:
  - Stale-read latency increases sharply.
  - Related TiDB metrics can show latency spikes around schema loading or `internal_others_leader`.
- Trigger / Preconditions:
  - Concurrent DDL activity is present.
  - Schema diffs are missing or cannot be applied incrementally, forcing more expensive schema reconstruction.
- Key metrics:
  - Schema-load related latency rises together with stale-read latency.
- Root cause:
  - Concurrent DDL can cause schema cache misses and missing schema diffs.
  - Stale-read queries then need to load snapshot schemas repeatedly, creating hotspots and latency spikes.
- Workaround:
  - Upgrade to a release that includes the fix associated with `tidb#53428`.
- Fixed version:
  - Refer to the `tidb#53428` release line before relying on version-only mitigation.
- Related oncall:
  - v6.5.4 `internal_others_leader` and schema load showed latency spikes `GTOC-6921`