Skip to main content
The xlog_location collector monitors the current transaction log (xlog) position on PostgreSQL servers.

​
Status

Default: Disabled Enable with: --collector.xlog-location
This collector is deprecated and only works on PostgreSQL 9.x. It will automatically skip collection on PostgreSQL 10+.In PostgreSQL 10, “xlog” was renamed to “wal”. Use the WAL-related collectors for PostgreSQL 10+.

​
Metrics

​
pg_xlog_location_bytes

Type: Gauge
Description: PostgreSQL LSN (Log Sequence Number) being generated on primary or replayed on replica Labels: None Note: The LSN value is truncated to the low 52 bits to fit in a float64.

​
SQL Query

SELECT CASE
  WHEN pg_is_in_recovery() THEN 
    (pg_last_xlog_replay_location() - '0/0') % (2^52)::bigint
  ELSE 
    (pg_current_xlog_location() - '0/0') % (2^52)::bigint
END AS bytes

​
How It Works

The collector:
  1. Checks if the server is in recovery (replica) or primary
  2. On primary: Returns current WAL write position (pg_current_xlog_location())
  3. On replica: Returns last replayed WAL position (pg_last_xlog_replay_location())
  4. Converts LSN to bytes and truncates to 52 bits

​
PostgreSQL Versions

Supported: PostgreSQL 9.x only PostgreSQL 10+ behavior: The collector detects PostgreSQL 10+ and logs a warning, skipping collection.

​
Why Truncate to 52 Bits?

LSN values in PostgreSQL are 64-bit integers, but Prometheus stores metrics as float64. To avoid precision loss, the value is truncated using modulo 2^52 (the mantissa size of float64). This is sufficient for monitoring rate of change but may wrap around after ~4.5 petabytes of WAL.

​
Required Permissions

The monitoring user needs:
  • Execute permission on pg_current_xlog_location() and pg_last_xlog_replay_location() functions
  • Granted to PUBLIC by default on PostgreSQL 9.x

​
Example Output

On Primary: 
pg_xlog_location_bytes 123456789012
On Replica: 
pg_xlog_location_bytes 123456780000

​
Use Cases (PostgreSQL 9.x Only)

​
Monitor WAL Generation Rate

# WAL bytes generated per second
rate(pg_xlog_location_bytes[5m])

​
Estimate Replication Lag

Combine primary and replica metrics: 
# Bytes behind primary (approximate)
pg_xlog_location_bytes{instance="primary"} - 
pg_xlog_location_bytes{instance="replica"}

​
Migration to PostgreSQL 10+

When upgrading from PostgreSQL 9.x to 10+:
  1. Disable this collector: 
    --no-collector.xlog-location
  2. Use replacement collectors:
    • WAL position monitoring: Built into replication collector
    • WAL file monitoring: Use wal collector
    • Replication lag: Use replication collector

​
Function Name Changes in PostgreSQL 10

PostgreSQL 9.xPostgreSQL 10+
pg_current_xlog_location()pg_current_wal_lsn()
pg_last_xlog_replay_location()pg_last_wal_replay_lsn()
pg_xlog/ directorypg_wal/ directory

​
Example Migration

PostgreSQL 9.x configuration: 
pgexporter \
  --collector.xlog-location \
  --web.listen-address=:9187
PostgreSQL 10+ configuration: 
pgexporter \
  --collector.wal \
  --collector.replication \
  --web.listen-address=:9187

​
Troubleshooting

​
Collector Skipped on PostgreSQL 10+

Expected behavior. The collector logs: 
xlog_location collector is not available on PostgreSQL >= 10.0.0, skipping
This is normal - use the wal and replication collectors instead.

​
Function Does Not Exist

On PostgreSQL 10+, you’ll see: 
ERROR: function pg_current_xlog_location() does not exist
Solution: Disable this collector for PostgreSQL 10+.

Build docs developers (and LLMs) love

Get started for freeTalk to us