Server Administration¶

📖 mongod Reference - mongod process documentation 📖 Configuration Options - Complete configuration reference

mongod Process¶

Starting mongod¶

Command Line:

mongod --dbpath /var/lib/mongodb --logpath /var/log/mongodb/mongod.log \
  --port 27017 --bind_ip 0.0.0.0 --fork

Configuration File:

mongod --config /etc/mongod.conf
# or
mongod -f /etc/mongod.conf

mongod.conf Structure¶

# Storage configuration
storage:
  dbPath: /var/lib/mongodb
  journal:
    enabled: true
    commitIntervalMs: 100
  directoryPerDB: true
  wiredTiger:
    engineConfig:
      cacheSizeGB: 4
      journalCompressor: snappy
      directoryForIndexes: true
    collectionConfig:
      blockCompressor: snappy
    indexConfig:
      prefixCompression: true

# System logging
systemLog:
  destination: file
  path: /var/log/mongodb/mongod.log
  logAppend: true
  logRotate: reopen
  verbosity: 0
  component:
    command:
      verbosity: 1

# Network
net:
  port: 27017
  bindIp: 0.0.0.0
  maxIncomingConnections: 65536
  tls:
    mode: requireTLS
    certificateKeyFile: /etc/ssl/mongodb.pem
    CAFile: /etc/ssl/ca.pem

# Process management
processManagement:
  fork: true
  pidFilePath: /var/run/mongodb/mongod.pid

# Security
security:
  authorization: enabled
  keyFile: /etc/mongodb/keyfile

# Replication
replication:
  replSetName: myReplicaSet
  oplogSizeMB: 10240

# Sharding
sharding:
  clusterRole: shardsvr  # or configsvr

# Profiling
operationProfiling:
  mode: slowOp
  slowOpThresholdMs: 100

Key Configuration Sections¶

WiredTiger Storage Engine¶

📖 WiredTiger - Storage engine documentation

Architecture¶

Key Components: - Internal Cache - Working set data in memory - Journal - Write-ahead log for durability - Checkpoints - Consistent snapshots of data on disk - Compression - On-disk data and index compression

Cache Management¶

Default Cache Size:

50% of (RAM - 1 GB)
Minimum: 256 MB

Configure Cache:

storage:
  wiredTiger:
    engineConfig:
      cacheSizeGB: 8

Cache Monitoring:

db.serverStatus().wiredTiger.cache
// Key metrics:
// "bytes currently in the cache"
// "maximum bytes configured"
// "pages read into cache"
// "pages written from cache"
// "tracked dirty bytes in the cache"
// "unmodified pages evicted"

Cache Pressure: - When dirty pages exceed 5% of cache, background eviction starts - When dirty pages exceed 20%, application threads assist eviction - Sudden cache pressure indicates working set exceeds cache size - Solution: increase cache size or optimize queries to use less data

Compression¶

Changing Compression:

// Per-collection compression
db.createCollection("myCollection", {
  storageEngine: {
    wiredTiger: {
      configString: "block_compressor=zstd"
    }
  }
});

• snappy - Fast compression, moderate ratio (default)
• zlib - Higher ratio, more CPU
• zstd - Best ratio at reasonable CPU
• none - No compression, fastest but most disk space

Checkpoints¶

• WiredTiger creates checkpoints every 60 seconds
• A checkpoint is a consistent snapshot of data on disk
• Journal entries between checkpoints ensure durability
• If mongod crashes, recovery replays journal from last checkpoint
• storage.syncPeriodSecs (default: 60) - Not typically changed

Journaling¶

📖 Journaling - Journaling documentation

• Write-ahead log (WAL) for crash recovery
• Enabled by default and should not be disabled in production
• Journal commit interval: 100 ms (default)
• With j: true write concern, wait for journal commit before acknowledging
• Journal files stored in journal/ subdirectory of dbPath
• Recovery: replay journal entries after last checkpoint

Journal Commit Interval:

storage:
  journal:
    commitIntervalMs: 100  # Default, reduce for lower durability gap

mongos Process¶

📖 mongos Reference - mongos documentation

# mongos.conf
sharding:
  configDB: configReplicaSet/config1:27019,config2:27019,config3:27019

net:
  port: 27017
  bindIp: 0.0.0.0

security:
  keyFile: /etc/mongodb/keyfile

systemLog:
  destination: file
  path: /var/log/mongodb/mongos.log

Key Differences from mongod: - No storage section (mongos does not store data) - No replication section - Must specify configDB connection string - Routes queries to appropriate shards - Multiple mongos instances for high availability

Database Profiler¶

📖 Database Profiler - Profiler documentation

Profiler Levels¶

Configuration¶

// Set profiler level
db.setProfilingLevel(1, { slowms: 100 });

// Check current profiling level
db.getProfilingStatus();
// { was: 1, slowms: 100, sampleRate: 1 }

// Set with sample rate (profile 50% of slow ops)
db.setProfilingLevel(1, { slowms: 100, sampleRate: 0.5 });

Querying Profile Data¶

// Find slow queries
db.system.profile.find({ millis: { $gt: 200 } })
  .sort({ ts: -1 }).limit(10);

// Find collection scans
db.system.profile.find({ "planSummary": "COLLSCAN" })
  .sort({ millis: -1 });

// Find queries on specific collection
db.system.profile.find({ ns: "mydb.mycollection" })
  .sort({ ts: -1 }).limit(5);

Profile Document Fields: | Field | Description | |-------|-------------| | op | Operation type (query, insert, update, etc.) | | ns | Namespace (database.collection) | | millis | Execution time in milliseconds | | ts | Timestamp | | planSummary | Query plan summary (IXSCAN, COLLSCAN) | | nreturned | Number of documents returned | | docsExamined | Number of documents examined | | keysExamined | Number of index keys examined | | command | The command that was profiled |

Server Diagnostic Commands¶

db.serverStatus()¶

const status = db.serverStatus();

// Connection info
status.connections
// { current: 42, available: 65494, totalCreated: 1500 }

// Operations counters
status.opcounters
// { insert: 1000, query: 5000, update: 2000, delete: 100, command: 8000 }

// Memory info
status.mem
// { bits: 64, resident: 4096, virtual: 8192 }

// WiredTiger cache
status.wiredTiger.cache
// { "bytes currently in the cache": ..., "maximum bytes configured": ... }

// Replication info
status.repl
// { hosts: [...], setName: "rs0", primary: "host:27017", me: "host:27017" }

db.currentOp()¶

// All current operations
db.currentOp();

// Long-running operations (> 5 seconds)
db.currentOp({ "secs_running": { $gt: 5 } });

// Operations on specific namespace
db.currentOp({ "ns": "mydb.mycollection" });

// Kill a long-running operation
db.killOp(opId);

db.stats() and Collection Stats¶

// Database stats
db.stats();
// { db: "mydb", collections: 10, objects: 50000, dataSize: ..., storageSize: ..., indexes: 25, indexSize: ... }

// Collection stats
db.collection.stats();
// { count: 10000, size: ..., storageSize: ..., nindexes: 3, totalIndexSize: ... }

// Collection stats with scale
db.collection.stats({ scale: 1048576 });  // In MB

Log Management¶

Log Configuration¶

systemLog:
  destination: file
  path: /var/log/mongodb/mongod.log
  logAppend: true
  logRotate: reopen  # or rename
  verbosity: 0
  component:
    accessControl:
      verbosity: 1
    command:
      verbosity: 0
    replication:
      verbosity: 1
    storage:
      verbosity: 0

Log Rotation¶

// Trigger log rotation
db.adminCommand({ logRotate: 1 });

With logRotate: reopen: - MongoDB closes and reopens the log file - External tool (logrotate) handles file rename - Send SIGUSR1 to rotate

With logRotate: rename: - MongoDB renames the file and opens a new one - Simpler but less flexible

Structured Logging (MongoDB 4.4+)¶

• Logs in JSON format by default
• Machine-parseable for log aggregation tools
• Component-based verbosity control
• Log messages have severity levels: F, E, W, I, D1-D5

Key Log Messages to Watch¶

• "msg":"Slow query" - Slow operations
• "msg":"Connection accepted" / "Connection ended" - Connection lifecycle
• "msg":"Transition" - Replica set state changes
• "msg":"WiredTiger" - Storage engine events
• "msg":"authentication" - Authentication attempts