System Design Patterns: From Fundamentals to Real Systems
Vote

0% completed

​

Write-Ahead Log

1. The Incident

Tuesday, 3:07 AM: a power failure shuts down the rack holding shard 2. The machine reboots ninety seconds later. Now one of two bad mornings begins.

Bad morning A: the database refuses to start. It reports "Data files corrupted." Here is how that happens. A single UPDATE is not one disk operation. Changing an order's status touches the table row and two indexes. That is three separate places on disk. The power died after the first write and before the other two. Now the files disagree with each other: an index points at a row that says something else

.....

.....

.....

Like the course? Get enrolled and start learning!

Reading Progress

0%