Locks & Waits

The Locks & Waits tab provides visibility into PostgreSQL concurrency behavior. Use it to identify lock contention, detect deadlocks, analyze blocking sessions, and troubleshoot wait events.

---

Sections

The Locks & Waits tab is organized into four sections:

1. Concurrency Overview — Lock types and deadlock monitoring
2. Locks Grid — Detailed view of all locks
3. Blocking Analysis — Blocking chains and impact assessment
4. Wait Event Analysis — PostgreSQL wait event distribution

---

Concurrency Overview Section

The Concurrency Overview section provides immediate visibility into lock activity.

Lock Count by Type

A pie chart showing the distribution of lock types.

Lock types:

Type	Description
relation	Table-level locks
tuple	Row-level locks
transactionid	Transaction ID locks
virtualxid	Virtual transaction ID locks
object	Database object locks
userlock	User-defined advisory locks
advisory	Advisory locks

Healthy pattern:

• Mostly relation and transactionid locks
• Few tuple locks (row-level contention)
• Advisory locks if using application locking

When to investigate:

• High tuple lock count — row contention
• Growing lock counts — possible lock escalation
• Unexpected lock types — review application behavior

Lock Wait Queue Depth

What it shows: Number of sessions waiting for locks over time.

Healthy range: Near zero for most workloads.

When to investigate:

• Queue depth > 0 sustained
• Growing trend
• Spikes correlating with slowness

High queue causes:

• Long-running transactions
• Lock contention on hot rows
• Missing indexes causing table locks

Deadlock Frequency

What it shows: Number of deadlocks over time.

Healthy range: Zero.

When to investigate:

• Any deadlocks occurring
• Recurring patterns
• Increasing frequency

Deadlock causes:

• Circular lock dependencies
• Inconsistent lock ordering
• Long transactions increasing collision probability

---

Locks Grid Section

The Locks Grid shows detailed information about all current locks. This section is collapsed by default — click to expand.

Table Columns

Column	Description
PID	Backend process ID
Database	Database name
Lock Type	Type of lock held/requested
Mode	Lock mode (AccessShare, RowExclusive, etc.)
Granted	Whether lock is held (true) or waiting (false)
Relation	Table or object being locked
Wait Start	When waiting began

Lock Modes

Mode	Description	Conflicts With
AccessShareLock	SELECT	AccessExclusiveLock
RowShareLock	SELECT FOR UPDATE	Exclusive, AccessExclusive
RowExclusiveLock	UPDATE, DELETE, INSERT	Share, ShareRowExclusive, Exclusive, AccessExclusive
ShareUpdateExclusiveLock	VACUUM, ANALYZE	ShareUpdateExclusive, Share, ShareRowExclusive, Exclusive, AccessExclusive
ShareLock	CREATE INDEX	RowExclusive, ShareUpdateExclusive, ShareRowExclusive, Exclusive, AccessExclusive
ShareRowExclusiveLock	Rarely used	RowExclusive, ShareUpdateExclusive, Share, ShareRowExclusive, Exclusive, AccessExclusive
ExclusiveLock	Blocks reads/writes	All except AccessShare
AccessExclusiveLock	ALTER TABLE, DROP	All

How to Use

Find blocked sessions:

• Filter for Granted = false
• These sessions are waiting for locks

Identify blockers:

• Find locks on same relation
• Match PIDs with Granted = true

Track wait duration:

• Check Wait Start timestamp
• Long waits indicate serious contention

---

Blocking Analysis Section

The Blocking Analysis section helps identify blocking chains and their impact. This section is collapsed by default — click to expand.

Blocking Sessions Node Graph

A visual representation of which sessions are blocking others.

What it shows:

• Blocking session → Blocked session relationships
• Chain of blocked processes
• Root blocker identification

How to use it:

• Identify the root cause of blocking
• See cascade effects of blocking
• Prioritize which session to address

Lock Escalation Pattern

What it shows: Lock escalation behavior over time.

Lock escalation:

• Row locks promoted to table locks
• Can cause widespread blocking
• Often indicates suboptimal queries

When to investigate:

• Frequent escalations
• Correlation with blocking
• After query changes

Impact Assessment

What it shows: Business impact of current blocking.

Metrics:

• Number of blocked sessions
• Total wait time
• Affected queries/operations

How to use it:

• Prioritize intervention
• Quantify blocking impact
• Track improvement

Advisory Locks

What it shows: Application-level advisory locks.

Advisory lock types:

• Session-level (released at session end)
• Transaction-level (released at transaction end)

When to investigate:

• Application using advisory locks
• Unexpected lock counts
• Locks not being released

---

Wait Event Analysis Section

The Wait Event Analysis section shows where PostgreSQL is spending time waiting. This section is collapsed by default — click to expand.

Wait Event Distribution

What it shows: Distribution of wait events by type.

Wait event types:

Type	Description
Client	Waiting for client (ClientRead, ClientWrite)
Lock	Waiting for heavyweight locks
LWLock	Waiting for lightweight locks
BufferPin	Waiting for buffer pin
IO	Waiting for I/O
Activity	Background worker activity
Extension	Extension-related waits

Healthy pattern:

• Mostly Client waits (waiting for queries)
• Low Lock waits
• Minimal IO waits

IO Wait Analysis

What it shows: I/O-related wait events.

IO wait types:

• DataFileRead — Reading data files
• DataFileWrite — Writing data files
• WALWrite — Writing WAL
• WALSync — Syncing WAL

When to investigate:

• High DataFileRead — buffer cache misses
• High WALWrite — heavy write workload
• Any sustained IO waits — storage bottleneck

LWLock Analysis

What it shows: Lightweight lock wait events.

Common LWLocks:

• buffer_content — Buffer access
• WALInsert — WAL insertion
• lock_manager — Lock management
• proc — Process management

When to investigate:

• High buffer_content — buffer contention
• High WALInsert — WAL write contention
• Unusual patterns — configuration issues

---

Use Cases

Diagnosing Lock Contention

When experiencing slow transactions:

1. Check Lock Wait Queue Depth for queueing
2. Review Lock Count by Type for lock patterns
3. Expand Locks Grid to see specific locks
4. Use Blocking Sessions Graph to find blockers
5. Check Impact Assessment for severity

Investigating Deadlocks

When deadlocks are occurring:

1. Monitor Deadlock Frequency for patterns
2. Enable log_lock_waits and deadlock_timeout
3. Review PostgreSQL logs for deadlock details
4. Analyze query patterns causing deadlocks
5. Implement consistent lock ordering

-- Enable deadlock logging
ALTER SYSTEM SET deadlock_timeout = '1s';
ALTER SYSTEM SET log_lock_waits = on;
SELECT pg_reload_conf();

Resolving Active Blocking

When you need to resolve blocking immediately:

1. Identify blocker in Blocking Sessions Graph
2. Check Impact Assessment for severity
3. Review blocker's query in Locks Grid
4. Decision:
   • Wait for completion
   • Cancel query: SELECT pg_cancel_backend(pid)
   • Terminate session: SELECT pg_terminate_backend(pid)

-- Cancel a query (graceful)
SELECT pg_cancel_backend(12345);

-- Terminate a session (forceful)
SELECT pg_terminate_backend(12345);

Performance Tuning for Concurrency

To improve concurrent access:

1. Review Wait Event Distribution for bottlenecks
2. Check LWLock Analysis for internal contention
3. Analyze IO Wait Analysis for I/O bottlenecks
4. Consider:
   • Reducing transaction duration
   • Optimizing query access patterns
   • Using appropriate isolation levels
   • Implementing row-level locking strategies

Monitoring Advisory Locks

If your application uses advisory locks:

1. Check Advisory Locks table for current state
2. Ensure locks are being released properly
3. Monitor for lock accumulation
4. Review application lock patterns

---

Related Metrics

The Locks & Waits section uses these metrics from the Metrics Reference:

Panel	Primary Metrics
Lock Count by Type	pg_locks_count
Lock Wait Queue	pg_locks_count (granted=false)
Deadlock Frequency	pg_database_stats.deadlocks
Locks Grid	pg_lock_detail, pg_backend_info
Blocking Sessions	pg_blocking_pids
Wait Events	pg_backend_wait_events

---

Related Guides

• Performance — Query performance analysis
• Connections — Connection management
• Queries — Query analysis