π The hard way.
It wasnβt from a course or documentation β it was from bringing down production.
Not because of a DELETE without WHERE, but from a massive delete that quietly filled up the transaction logs and exhausted every bit of allocated space.
Production froze. Everyone looked at the DB.
And the culprit? Me. π
Iβve done it twice β once on Sybase (T-log) and once on Oracle (Redo Log).
Different companies, same mistake.
So hereβs my message to all the data analysts and data wranglers:
If youβre using the production database to temporarily store and later delete large analysis datasets β be aware that those deletes are fully logged.
Every delete = undo + redo + log space consumed.
If the log fills up, the whole system can grind to a halt.
β
Use NOLOGGING / minimal logging where possible
β
Do batched deletes with commits
β
Or better yet β donβt use prod as your scratchpad
Your DBA (and your uptime) will thank you. π
Top comments (0)