MongoDB Timeout & Transaction Reproduction

Note: This is deprecated. Use V2.

Local 3-member MongoDB replica set for testing timeouts, transactions, and failure scenarios.

Setup

Requirements: Docker, Node.js 18+, mongosh

1. Add hostnames to /etc/hosts:

echo "127.0.0.1 mongo1 mongo2 mongo3" | sudo tee -a /etc/hosts

2. Start replica set:

npm install
npm run start:db
npm run init:rs

3. Build and run test:

npm run build
npm run run:tx

Connection string: mongodb://mongo1:27017,mongo2:27017,mongo3:27017/?replicaSet=rs0

Test Scenarios

Regular Test (With Timeouts - Throws Errors)

Lose majority (WriteConcernFailed):

npm run stop:member -- mongo2
npm run stop:member -- mongo3
npm run run:tx
npm run start:member -- mongo2
npm run start:member -- mongo3

Force stepdown:

npm run stepdown
npm run run:tx

Kill member (crash simulation):

npm run kill:member -- mongo2
npm run run:tx
npm run start:member -- mongo2

Operation timeout:

export OP_MAXTIME_MS=1
npm run run:tx

Production Simulation (NO Timeouts - Hangs Indefinitely)

⚠️ WARNING: These tests have NO timeouts and will HANG INDEFINITELY if majority is lost (simulating production behavior). Use Ctrl+C to kill the process.

Option A: Hang on Write Operation

1. Start the no-timeout test:

npm run build
npm run run:tx-notimeout

2. When you see "💡 STOP NODES NOW! You have 5 seconds...", quickly run:

docker stop mongo2 mongo3

3. The script should hang indefinitely at the update operation.

4. Use Ctrl+C to kill, then restore:

docker start mongo2 mongo3

Option B: Hang on Commit (Most Common Production Scenario)

This reproduces the most common production case where operations complete but commit hangs:

1. Start the hang-on-commit test:

npm run build
npm run run:tx-hangcommit

2. Wait for the countdown (20 seconds), then when prompted, choose ONE of these:

   Option 1: Network Partition (More Realistic)

   npm run partition:member -- mongo2 mongo3

   This simulates a network partition where nodes stay running but become unreachable (closer to production behavior).

   Option 2: Stop Containers (Simpler)

   docker stop mongo2 mongo3

   This is simpler but the primary detects the failure faster and may throw an error instead of hanging.

3. The script should hang indefinitely at commitTransaction()!

4. Use Ctrl+C to kill, then restore:

   # If you used network partition:
   npm run restore:member -- mongo2 mongo3
   
   # If you used docker stop:
   docker start mongo2 mongo3

Key differences in no-timeout tests:

• socketTimeoutMS=0 (infinite socket timeout)
• No wtimeout in write concern
• No maxCommitTimeMS on transaction
• No maxTimeMS on operations

This matches typical production configurations where timeouts are not explicitly set, causing commits to hang forever when replica set loses majority.

Configuration

Environment variables (defaults shown):

W_TIMEOUT_MS=2000       # Write concern timeout
MAX_COMMIT_MS=1500      # Transaction commit timeout
OP_MAXTIME_MS=1000      # Operation timeout

Cleanup

docker compose down -v