Diagnostics

Diagnose local-ydb with MCP tools

Start broad and read-only, then route by YDB healthcheck issue type. Do not repair a local-ydb stack before collecting current evidence.

Recommended order

Run local_ydb_check_prerequisites on a new host or profile.
Run local_ydb_status_report to capture Docker, tenant, node, auth, and health context.
Run local_ydb_healthcheck with noCache=true when a fresh YDB self-check signal is needed. Call the result fresh only if noCache remains in optionResolution.effective, and inspect warnings and compatibilityFallback.
Route by issue type rather than trying repair commands immediately.

Route by symptom

STORAGE: run local_ydb_storage_placement, then inspect static and dynamic logs.
COMPUTE, COMPUTE_POOL, tablet, or node issues: run local_ydb_nodes_check, local_ydb_tenant_check, and dynamic container logs.
DATABASE or SCHEME symptoms: run local_ydb_database_status and local_ydb_scheme.
Auth symptoms: run local_ydb_auth_check.

Prompt support

The MCP server includes local_ydb_diagnose_stack and local_ydb_diagnose_database prompts. They instruct the client to inspect current state, summarize observed evidence, and avoid automatic repair.