Diagnostic query, accessing DATA_LEGTH or INDEX_LENGTH from information_schema.TABLES table which runs in parallel with long ANALYZE TABLE crashes the server. This can be a query, accessing information_schema.TABLES directly as well as a query like SHOW TABLE STATUS.
While this crash is intetional both operations are valid, there is no deadlock. If ANALYZE TABLE runs slightly less than 600 seconds both queries finish successfully.
How to repeat:
1. Create a table with indexes big enough to ANALYZE TABLE to run more than 600 seconds. Use large value for STATS_SAMPLE_PAGES
2. Start ANALYZE TABLE
3. In parallel connection run
4. Wait for slightly more than 600 seconds and observe the crash.
Do not crash.
ha_innobase::info_low accepts flag HA_STATUS_NO_LOCK Maybe call such a diagnostic query with this flag?