Deadlock in DBMS

Deadlock kya hota hai?

DBMS mein deadlock ek aisi situation hoti hai jahan 2 ya zyada transactions ek dusre ka wait karte rehte hain resource release krne ke liye, aur koi bhi aage nahi badh pata.

example:

Imagine:

Person A ke pas pen hai
Person B ke pas notebook hai
A bolta hai “Mai sirf tabhi hi likhunga jab mjhe notebook milegi”
B bolta hai “Mai tumhe notebook tabhi hi dunga jab mujhe pen milegi”

Koi bhi nahi likh pata na hi A, na hi B .ise hi deadlock kehte hai.

Deadlock ko handle karna kyun zaroori hai?

Agar hum deadlock ko handle nahi karte:

Transactions indefinitely wait karte rahenge
System hang ya slow ho sakta hai
Performance aur data consistency kharaab ho jaati hai

Deadlock ko handle karne ke 4 main tareeke

1. Deadlock Prevention

Goal: Deadlock hone hi na do. Isme hum 4 Coffman Conditions me se koi ek tod dete hain.

Coffman Conditions:

Condition	Matlab
Mutual Exclusion	Ek resource ek waqt mein sirf ek hi le sakta hai (resource non sharable hai)
Hold and Wait	Transaction ek resource pakad ke dusra maangta hai
No Preemption	Resource zabardasti nahi cheen sakte
Circular Wait	Transactions ek dusre ka wait kar rahe hote hain in cycle

Kaise rok sakte hain?

a) Hold and Wait todna

Transaction ko bolte hain ki saare resources ek saath maango. Agar mil gaye to chalo, warna wait karo.

Example:
T1 ko A aur B chahiye — dono ek saath maange, nahi to wait kare.

b) Circular Wait todna

Resources ko number de do (e.g. A=1, B=2, C=3). Har transaction ko increasing order mein hi resource maangna hoga.

Example:
T1 ne A (1) le liya — ab sirf B (2) ya C (3) maang sakta hai, A se chhota resource nahi.

2. Deadlock Avoidance

Goal: Deadlock hone se pehle hi check karo ki situation safe hai ya nahi.

Safe State:

Aisi condition jahan har transaction safely complete ho sakta hai.

Unsafe State:

Aisi condition jahan aage jaake deadlock ho sakta hai.

Algorithm: Banker’s Algorithm

Example:
T1 resource A maangta hai → system check karega:

Kya sabhi transactions safe tareeke se complete ho sakte hain?
Agar haan → resource de do
Agar nahi → wait karvao

Iske liye har transaction ka maximum resource requirement pehle se pata hona chahiye

3. Deadlock Detection & Recovery (Pehchaan ke baad sudharna)

Goal: Deadlock hone do, par turant detect karo aur solve karo.

Kaise detect karte hain?

Use karte hain Wait-For Graph (WFG):

Har node = ek transaction
Edge = T1 is waiting for T2

Cycle in graph = Deadlock

Example Wait-for Graph

Ye cycle hai ⇒ deadlock hua

Recovery kaise karte hain?

Deadlock detect hone ke baad system kisi ek transaction ko rollback kar deta hai:

Recovery options:

Sabse young transaction ko kill karo
Sabse purana transaction ko rollback karo
Least costly (kam resources) transaction rollback karo
Partial rollback bhi kar sakte ho

Example:
T1, T2, T3 deadlock mein hain. T2 ko rollback kar diya, uske resources free ho gaye, aur baaki aage badh gaye.

4. Deadlock Ignorance (Ignore karna)

Kuch DBMS (jaise MySQL) deadlock ko ignore karte hain:

Sochte hain deadlock rare hota hai
User ko manually transaction retry karna padta hai

Ye approach serious systems ke liye safe nahi hai

Comparison Table

Method	Kab Use Hota Hai	Kaise Kaam Karta Hai	Example
Prevention	Pehle se	Deadlock ki condition tod deta hai	Saare resources ek saath maangna
Avoidance	Resource allocation se pehle	Safe/Unsafe state check karta hai	Banker’s Algorithm
Detection & Recovery	Deadlock hone ke baad	Cycle detect karta hai, rollback karta hai	Wait-for Graph
Ignore	Small systems mein	Khaali chhod dete hain, user handle kare	Manual retry

Final Summary

Deadlock tab hota hai jab transactions ek dusre ka wait karte hain aur koi bhi aage nahi badh pata
Hum deadlock ko prevent, avoid, detect or ignore kar sakte hain
Har method ka use depend karta hai system ki complexity, performance aur reliability pe

The Hingineer