note.md

How to prevent deadlock in golang

-> What is deadlock caused by ? Supposed 2 transactions occurs concurrently in same phrase => Transaction A is trigger update queries, same time transaction B is also triggers a transaction updates -> Transaction B will waiting the update results from transaction A in periods of waiting time till the operation response. <-> Solutions :

• Using update query by order

SELECT ... FOR UPDATE (1)

• Locking mechanisms used within transaction
• Prevent race conditions and ensure data consistency.

SELECT ... FOR NO KEY UPDATE (2)

• equivalence as (1) query -> but weaker
• selected rows are not modified in ways that affect the non-key columns ============================================================================

Why context in golang ?

• Carry deadlines, Cancellation signals, request-scoped data.
• Helps to propagating cancellation signals throughout the application.
• Initialize context by using:
  • context.WithCancel(parent) -> cancel by using cancel()
  • context.WithDeadline(parent, deadline) -> automatically canceled when deadline is reached.
  • context.WithTimeOut(parent, timeout) -> similarly to withDeadline
  • context.WithValue(parent, key, value) -> [often use with the request-scoped] ============================================================================

Dives deeper into Transaction

Command showing the current isolation levels : - show transaction isolation level Command options for setting isolation levels : - set session transaction isolation level serializable

1. Read Phenomena

• Dirty Read: transaction reads data written by other concurrent uncommitted transaction
• Non-repeated read: reads the same row twice and sees difference value
• Phantom read : transaction re-executes a query to find rows satisfy the condition and sees the difference set of rows
• Serialization anomaly :

2. 4 standards isolation levels

• Read uncommitted -> read committed -> Repeatable read -> Serializable read