# Feishu Storage Oncall Handbook to Skill-Family Mapping This document records how the Feishu handbook content was normalized into the current TiKV oncall skill family. ## Source Documents ### Primary root handbook - `Storage oncall handbook` - Feishu: - Role in skill family: source umbrella only, not a standalone skill ### Related child documents reviewed during migration - `Raftstore handbook` - Feishu: - Role in skill family: incremental source for `tikv-raftstore` and `tikv-performance` - `Understanding TiKV Memory Management and common issues` - Feishu: - Role in skill family: incremental source for `tikv-memory` - `Transaction oncall handbook` - Feishu: - Role in skill family: already represented by `transaction-oncall-handbook` ## Target Skill Family - `skills/diagnosis/tikv-oncall-handbook/SKILL.md` - `skills/diagnosis/tikv-panic/SKILL.md` - `skills/diagnosis/tikv-raftstore/SKILL.md` - `skills/diagnosis/tikv-recovery/SKILL.md` - `skills/diagnosis/tikv-performance/SKILL.md` - `skills/diagnosis/tikv-storage/SKILL.md` - `skills/diagnosis/tikv-memory/SKILL.md` - `skills/diagnosis/tikv-status-server/SKILL.md` - `skills/diagnosis/tikv-cdc/SKILL.md` - `skills/diagnosis/transaction-oncall-handbook/SKILL.md` ## Mapping Rules Used - Keep one entry skill: `tikv-oncall-handbook` - Split by incident family, not by original document boundaries - Keep main `SKILL.md` files short and routing-oriented - Move long case detail into `references/*.md` - Use English only for normalized skill content - Remove emoji-based status markers from the main oncall entry skills - Avoid duplicate long-form explanations when a sibling skill already owns the topic ## Root Handbook Coverage ### Root handbook topics routed to `tikv-raftstore` - add-learner timeout loops - delete-snapshot backlog on offline node - ingest SST during data movement - startup apply-busy delaying leader scheduling - stale-peer overlap and snapshot/apply side effects Implemented in: - `skills/diagnosis/tikv-raftstore/SKILL.md` - `skills/diagnosis/tikv-raftstore/references/add-learner-timeout-loop.md` - `skills/diagnosis/tikv-raftstore/references/ingest-sst-during-data-movement.md` - `skills/diagnosis/tikv-raftstore/references/startup-apply-busy-and-leader-scheduling.md` ### Root handbook topics routed to `tikv-panic` - SIGSEGV on ARM weak-memory platforms - missing SST file panic - abnormal restart damages raft state - restart or upgrade failure cases - PITR pause or resume panic Implemented in: - `skills/diagnosis/tikv-panic/SKILL.md` - `skills/diagnosis/tikv-panic/references/sigsegv-on-arm-rocksdb.md` - `skills/diagnosis/tikv-panic/references/restart-and-upgrade-known-cases.md` ### Root handbook topics routed to `tikv-storage` - THP-related latency and memory regression - Titan GC and Titan size-estimation issues - host and disk behavior cases - compaction and SST-focused storage pressure Implemented in: - `skills/diagnosis/tikv-storage/SKILL.md` - `skills/diagnosis/tikv-storage/references/thp-latency-and-memory-regression.md` - `skills/diagnosis/tikv-storage/references/titan-behavior-known-cases.md` - `skills/diagnosis/tikv-storage/references/host-and-disk-behavior-known-cases.md` ### Root handbook topics routed to `tikv-status-server` - TLS mismatch on status port - pprof deadlock - profile collection failures on custom builds - profile FD leak - slow host-side backtrace collection Implemented in: - `skills/diagnosis/tikv-status-server/SKILL.md` - `skills/diagnosis/tikv-status-server/references/status-server-known-cases.md` ### Root handbook topics routed to `tikv-cdc` - TiKV address reuse causes CDC lag - stale PD endpoint reuse - CDC-triggered TiKV panic Implemented in: - `skills/diagnosis/tikv-cdc/SKILL.md` - `skills/diagnosis/tikv-cdc/references/tikv-address-reuse-changefeed-lag.md` ### Root handbook topics routed to `transaction-oncall-handbook` - client-go store-cache inconsistency - follower resolved-ts gap caused by stale peer - stale-read latency spikes due to schema loading - existing transaction and stale-read handbook content from sibling Feishu material Implemented in: - `skills/diagnosis/transaction-oncall-handbook/SKILL.md` - `skills/diagnosis/transaction-oncall-handbook/references/client-go-store-pointer-inconsistency.md` - `skills/diagnosis/transaction-oncall-handbook/references/follower-resolved-ts-gap-stale-peer.md` - `skills/diagnosis/transaction-oncall-handbook/references/stale-read-latency-heavy-schema-loading.md` ## Child-Document Coverage Added After Root Migration ### `Raftstore handbook` Additional cases normalized into: - `tikv-performance` - raft admin bursts and write-pipeline stalls - split or merge I/O blocking write pipeline - false slow-store judgment under high write rate - `tikv-raftstore` - startup apply-busy and delayed leader scheduling ### `Understanding TiKV Memory Management and common issues` Additional cases normalized into: - `tikv-memory` - resolved-ts OOM patterns - swap-related failures and jitter - heap-profile collection and configuration guidance - pessimistic-lock and raft-router memory leak cases ## Intentionally Not Duplicated These topics were not copied again if a sibling skill already owned them: - transaction correctness and stale-read logic already covered by `transaction-oncall-handbook` - CDC endpoint reuse already covered by `tikv-cdc` - generic compaction and SST pressure already covered by `tikv-storage` - region-lifecycle and split overlap cases already covered by `tikv-raftstore` ## Completion Notes This mapping is intended as an operational coverage document, not a line-by-line mirror of the Feishu pages. Completion standard used here: - the main incident families from the Feishu root handbook are represented in the skill family - high-frequency known cases were normalized into English - entry routing was tightened so an oncall engineer or agent can pick the right skill quickly - duplicated long-form explanations were reduced in favor of cross-links and references