Menu

Transactions in DBMS: Fast Revision Notes for IBPS SO IT Officer 2026

Jitendra Chadar
July 13, 2026
7 min read
DBMS
Transactions in DBMS: Fast Revision Notes for IBPS SO IT Officer 2026

Transactions in DBMS: Fast Revision Notes for IBPS SO IT Officer

A transaction in DBMS is a sequence of one or more database operations executed as a single logical unit of work, which either completes entirely or leaves the database unchanged. IBPS SO IT Officer and similar exams test this topic through conceptual questions on transaction states, COMMIT, ROLLBACK, and SAVEPOINT — rarely through pure definitions — so understanding the state transitions matters more than memorizing terms.

Questions on transaction states, COMMIT vs ROLLBACK, and SAVEPOINT are asked far more often than theoretical definitions. Master the state diagram below first.

What Is a Transaction?

A transaction is considered successful only when every operation inside it completes; if even one operation fails, the database rolls back to its previous consistent state.

A transaction typically groups operations like:

  • INSERT
  • UPDATE
  • DELETE
  • SELECT (in some cases, for read consistency)

Example: Fund Transfer

Rahul transfers ₹5,000 from Account A to Account B. The transaction has two operations:

  1. Debit ₹5,000 from Account A (₹50,000 → ₹45,000)
  2. Credit ₹5,000 to Account B (₹20,000 → ₹25,000)

Q: What happens if the debit succeeds but the credit operation fails? A: The DBMS rolls back the debit operation as well, since a transaction must be all-or-nothing — Account A's balance is restored to ₹50,000 so the database never ends up in an inconsistent, partially-updated state.

Why Transactions Are Needed

Without transactions, a partial update caused by any of the following could leave the database inconsistent:

CauseRisk Without Transactions
Power failureUpdate stops midway, data left inconsistent
System crashPartially written changes corrupt records
Concurrent accessTwo users overwrite each other's changes
Constraint violationInvalid data gets partially inserted

Transaction Lifecycle & States

Every transaction moves through six possible states — Active, Partially Committed, Committed, Failed, Aborted, and Terminated — depending on whether it completes successfully or encounters an error.

Click to enlarge
DBMS Transaction Lifecycle: Active to Partially Committed to Committed, or Failed to Aborted
The transaction state diagram is one of the most frequently asked visuals in DBMS exams — memorize both the success path and the failure path.

Success path: Active → Partially Committed → Committed → Terminated Failure path: Active → Failed → Aborted → Terminated (or restarted)

StateWhat's Happening
ActiveSQL statements are currently executing; the transaction is in progress
Partially CommittedThe last statement has executed, but changes aren't yet permanent
CommittedAll changes are permanently saved to disk; cannot be undone
FailedAn error occurred (constraint violation, deadlock, crash) before completion
AbortedThe DBMS has rolled back all changes made by the failed transaction
TerminatedFinal state — reached after either a successful COMMIT or a successful ROLLBACK

Q: Which transaction state occurs immediately after the last SQL statement executes, before COMMIT? A: Partially Committed. The statement has run, but the DBMS hasn't yet guaranteed the changes are permanently written to disk — that guarantee only comes after COMMIT succeeds.

Q: What is the difference between the Failed and Aborted states? A: Failed is the state entered the moment an error occurs (e.g., a constraint violation or deadlock), before any recovery happens. Aborted is the state entered after the DBMS has rolled back all the transaction's changes in response to that failure.

Q: Can a transaction resume after entering the Aborted state? A: A transaction cannot resume itself, but the DBMS or application may choose to restart it as a brand-new transaction; the original aborted transaction still ends in the Terminated state.

Transaction Control Commands (TCL)

COMMIT, ROLLBACK, SAVEPOINT, and ROLLBACK TO SAVEPOINT are the four Transaction Control Language (TCL) commands used to manage transaction outcomes, with COMMIT and ROLLBACK the most frequently tested.

COMMIT

COMMIT permanently saves all changes made during the current transaction, after which the changes become visible to other users and can no longer be undone.

START TRANSACTION;

UPDATE Employee
SET Salary = Salary + 5000
WHERE EmpID = 101;

COMMIT;

ROLLBACK

ROLLBACK cancels a transaction and restores the database to its state before the transaction began, but only for changes that have not yet been committed.

START TRANSACTION;

DELETE FROM Employee
WHERE EmpID = 101;

ROLLBACK;

The deleted record is restored, because the DELETE was never committed.

SAVEPOINT and ROLLBACK TO SAVEPOINT

A SAVEPOINT marks an intermediate checkpoint inside a transaction, letting you undo only the operations after that checkpoint instead of the entire transaction.

START TRANSACTION;

UPDATE Employee
SET Salary = Salary + 5000
WHERE EmpID = 101;

SAVEPOINT S1;

DELETE FROM Employee
WHERE EmpID = 102;

ROLLBACK TO S1;

COMMIT;

Result: the salary update is preserved, the delete is undone, and the transaction is committed successfully — this partial-rollback behavior is exactly what SAVEPOINT is tested on.

TCL Commands Summary

CommandPurpose
COMMITPermanently saves all changes
ROLLBACKUndoes all uncommitted changes
SAVEPOINTCreates an intermediate checkpoint
ROLLBACK TO SAVEPOINTUndoes only changes made after a specific savepoint

Q: Can ROLLBACK undo a transaction that has already been committed? A: No. Once COMMIT has executed successfully, the changes are permanent and ROLLBACK has no effect on them — ROLLBACK can only undo uncommitted work.

Autocommit Mode

Autocommit is a DBMS setting that, when ON, treats every individual SQL statement as its own transaction and commits it automatically right after execution.

Autocommit StateBehavior
ONEach statement is its own transaction; auto-committed immediately
OFFMultiple statements can be grouped; saved only when COMMIT runs manually

In competitive exams, assume transactions are executed manually (Autocommit OFF) unless the question explicitly says otherwise.

Transactions vs ACID Properties

Transactions define how database operations are grouped and executed, while ACID properties — Atomicity, Consistency, Isolation, and Durability — define the rules that make those transactions reliable.

TransactionsACID Properties
Logical unit of workGuarantees that make transactions trustworthy
Managed using TCL commandsIncludes Atomicity, Consistency, Isolation, Durability
Contains the actual SQL operationsDefines correctness rules the operations must satisfy

👉 Continue your revision with ACID Properties in DBMS — the two topics are frequently tested together in IBPS SO IT Officer.

One-Minute Revision

ConceptRemember
TransactionLogical unit of work — all or nothing
COMMITPermanently saves changes
ROLLBACKUndoes uncommitted changes
SAVEPOINTCheckpoint for partial rollback
Success pathActive → Partially Committed → Committed → Terminated
Failure pathActive → Failed → Aborted → Terminated
Committed transactionsCannot be rolled back
Autocommit ON (exam default: assume OFF)Each statement auto-commits individually

🎯 Revision Checklist

Tick the concepts you're confident about before attempting the mock test.

0/7
Revision Progress0%
Keep revising.

What to Study Next

Complete your DBMS revision with these related articles.

TopicStatus
SQL Commands✅ Available
SQL Constraints✅ Available
SQL Joins✅ Available
SQL Keys✅ Available
Normalization✅ Available
ACID Properties🚧 Coming Soon
Free Practice

Practice Transactions in DBMS

Test your understanding of transaction states, COMMIT, ROLLBACK, SAVEPOINT, and other DBMS concepts with exam-level questions.

Frequently Asked Questions (FAQs)

A transaction is a sequence of one or more database operations, such as INSERT, UPDATE, or DELETE, executed together as a single logical unit of work that either completes entirely or has no effect at all.

Ready to Practice?

Don't just read about it. Create a custom mock test for this topic instantly.