NoSQL to Distributed SQL: Best Practices and Lessons Learned by Helpshift
summary
Helpshift migrates MongoDB to a distributed SQL (yoga by db) by normalizing core fields into three tables issue, issue_chatbot_state, issue_meta while keeping volatile attributes in a single JSON `extra` field to avoid exploding schema work. Migration uncovers practical pitfalls: null characters in text, bind-variable limits on bulk inserts, and large secondary-index hotspots. Resolution includes hashing tenant IDs with yb_hash_code and a mod value to spread shards, favoring index-only scans and parallelizing large IN or aggregate queries into per-shard ranges.
AI-written summary · about
chapters (9)
- 0:00Migration journey intro
- 1:49Helpshift platform overview
- 3:34Switching from NoSQL to SQL
- 5:48NoSQL to SQL migration challenges
- 9:15Normalization in data modeling
- 15:33Data migration and schema changes
- 18:23Query patterns and indexing
- 24:29Anti-patterns in SQL queries
- 28:09Key takeaways
related talks
29mHow Helpshift Accelerated Growth by Migrating to
29mModern Database Architecture: From SQL to NoSQL & Distributed SQL
43mHelp! I Need To UnSQLize My Application
26mNewSQL: concepts, realities and when you should embrace it
30mHelp! I Need To UnSQLize My Application
40mData Migration Lessons from the Trenches
18mFrom SQL to NOSQL - a gentle intro for Devs
27mSparking our way to YugabyteDB: Our journey from ScyllaDB to