Logging v1

Suggest edits

The operator is designed to log in JSON format directly to standard output, including PostgreSQL logs.

Each log entry has the following fields:

  • level – Log level (info, notice, ...).
  • ts – The timestamp (epoch with microseconds).
  • logger – The type of the record (for example, postgres or pg_controldata).
  • msg – The actual message or the keyword record in case the message is parsed in JSON format.
  • record – The actual record with structure that varies depending on the logger type.
  • logging_podName – The pod where the log was created.
Warning

Long-term storage and management of logs is outside the operator's purview, and needs to be provided at the level of the Kubernetes installation. See the Kubernetes Logging Architecture documentation.

Info

If your log ingestion system requires it, you can rename the level and ts field names using the log-field-level and log-field-timestamp flags of the operator controller. Edit the Deployment definition of the cloudnative-pg operator.

Operator log

You can specify a log level in the cluster spec with the option logLevel. You can set it to error, warning, info(default), debug, or trace.

Currently, you can set the log level only when an instance starts. You can't change it at runtime. If you change the value in the cluster spec after the cluster was started, it takes effect only in the new pods and not the old ones.

PostgreSQL log

Each entry in the PostgreSQL log is a JSON object having the logger key set to postgres and the structure described in the following example:

{
  "level": "info",
  "ts": 1619781249.7188137,
  "logger": "postgres",
  "msg": "record",
  "record": {
    "log_time": "2021-04-30 11:14:09.718 UTC",
    "user_name": "",
    "database_name": "",
    "process_id": "25",
    "connection_from": "",
    "session_id": "608be681.19",
    "session_line_num": "1",
    "command_tag": "",
    "session_start_time": "2021-04-30 11:14:09 UTC",
    "virtual_transaction_id": "",
    "transaction_id": "0",
    "error_severity": "LOG",
    "sql_state_code": "00000",
    "message": "database system was interrupted; last known up at 2021-04-30 11:14:07 UTC",
    "detail": "",
    "hint": "",
    "internal_query": "",
    "internal_query_pos": "",
    "context": "",
    "query": "",
    "query_pos": "",
    "location": "",
    "application_name": "",
    "backend_type": "startup"
  },
  "logging_pod": "cluster-example-1",
}

Internally, the operator relies on the PostgreSQL CSV log format. See the PostgreSQL documentation for more information about the CSV log format.

PGAudit logs

EDB Postgres for Kubernetes has transparent and native support for PGAudit on PostgreSQL clusters.

To enable this support, add the required pgaudit parameters to the postgresql section in the configuration of the cluster.

Important

You need to add the PGAudit library to shared_preload_libraries. EDB Postgres for Kubernetes adds the library based on the presence of pgaudit.* parameters in the postgresql configuration. The operator detects and manages the addition and removal of the library from shared_preload_libraries.

The operator also takes care of creating and removing the extension from all the available databases in the cluster.

Important

EDB Postgres for Kubernetes runs the CREATE EXTENSION and DROP EXTENSION commands in all databases in the cluster that accept connections.

This example shows a PostgreSQL 13 Cluster deployment that results in pgaudit being enabled with the requested configuration:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-example
spec:
  instances: 3
  imageName: quay.io/enterprisedb/postgresql:13

  postgresql:
    parameters:
      "pgaudit.log": "all, -misc"
      "pgaudit.log_catalog": "off"
      "pgaudit.log_parameter": "on"
      "pgaudit.log_relation": "on"

  storage:
    size: 1Gi

The audit CSV logs entries returned by PGAudit are then parsed and routed to stdout in JSON format, similarly to all the remaining logs:

  • .logger is set to pgaudit.
  • .msg is set to record.
  • .record contains the whole parsed record as a JSON object. This is similar to logging_collector logs, except for .record.audit, which contains the PGAudit CSV message formatted as a JSON object.

This example shows sample log entries:

{
  "level": "info",
  "ts": 1627394507.8814096,
  "logger": "pgaudit",
  "msg": "record",
  "record": {
    "log_time": "2021-07-27 14:01:47.881 UTC",
    "user_name": "postgres",
    "database_name": "postgres",
    "process_id": "203",
    "connection_from": "[local]",
    "session_id": "610011cb.cb",
    "session_line_num": "1",
    "command_tag": "SELECT",
    "session_start_time": "2021-07-27 14:01:47 UTC",
    "virtual_transaction_id": "3/336",
    "transaction_id": "0",
    "error_severity": "LOG",
    "sql_state_code": "00000",
    "backend_type": "client backend",
    "audit": {
      "audit_type": "SESSION",
      "statement_id": "1",
      "substatement_id": "1",
      "class": "READ",
      "command": "SELECT FOR KEY SHARE",
      "statement": "SELECT pg_current_wal_lsn()",
      "parameter": "<none>"
    }
  },
  "logging_pod": "cluster-example-1",
}

See the PGAudit documentation for more details about each field in a record.

EDB Audit logs

Clusters that are running on EDB Postgres Advanced Server (EPAS) can enable EDB Audit as follows:

apiVersion: postgresql.k8s.enterprisedb.io/v1
kind: Cluster
metadata:
  name: cluster-example
spec:
  instances: 3
  imageName: quay.io/enterprisedb/edb-postgres-advanced:13
  licenseKey: <LICENSE>

  postgresql:
    epas:
      audit: true

  storage:
    size: 1Gi

Setting .spec.postgresql.epas.audit: true enforces the following parameters:

edb_audit = 'csv'
edb_audit_destination = 'file'
edb_audit_directory = '/controller/log'
edb_audit_filename = 'edb_audit'
edb_audit_rotation_day = 'none'
edb_audit_rotation_seconds = '0'
edb_audit_rotation_size = '0'
edb_audit_tag = ''
edb_log_every_bulk_value = 'false'

Other parameters can be passed via .spec.postgresql.parameters as usual.

The audit CSV logs are parsed and routed to stdout in JSON format, similarly to all the remaining logs:

  • .logger set to edb_audit
  • .msg set to record
  • .record containing the whole parsed record as a JSON object

See the example below:

{
  "level": "info",
  "ts": 1624629110.7641866,
  "logger": "edb_audit",
  "msg": "record",
  "record": {
    "log_time": "2021-06-25 13:51:50.763 UTC",
    "user_name": "postgres",
    "database_name": "postgres",
    "process_id": "68",
    "connection_from": "[local]",
    "session_id": "60d5df76.44",
    "session_line_num": "5",
    "process_status": "idle in transaction",
    "session_start_time": "2021-06-25 13:51:50 UTC",
    "virtual_transaction_id": "3/93",
    "transaction_id": "1183",
    "error_severity": "AUDIT",
    "sql_state_code": "00000",
    "message": "statement: GRANT EXECUTE ON function pg_catalog.pg_read_binary_file(text) TO \"streaming_replica\"",
    "detail": "",
    "hint": "",
    "internal_query": "",
    "internal_query_pos": "",
    "context": "",
    "query": "",
    "query_pos": "",
    "location": "",
    "application_name": "",
    "backend_type": "client backend",
    "command_tag": "GRANT",
    "audit_tag": "",
    "type": "grant"
  },
  "logging_pod": "cluster-example-1",
}

See EDB Audit file for more details about the records' fields.

Other logs

All logs that are produced by the operator and its instances are in JSON format, with logger set according to the process that produced them. Therefore, all the possible logger values are the following:

  • barman-cloud-wal-archive: from barman-cloud-wal-archive directly
  • barman-cloud-wal-restore: from barman-cloud-wal-restore directly
  • edb_audit: from the EDB Audit extension
  • initdb: from running initdb
  • pg_basebackup: from running pg_basebackup
  • pg_controldata: from running pg_controldata
  • pg_ctl: from running any pg_ctl subcommand
  • pg_rewind: from running pg_rewind
  • pgaudit: from PGAudit extension
  • postgres: from the postgres instance (having msg different than record)
  • wal-archive: from the wal-archive subcommand of the instance manager
  • wal-restore: from the wal-restore subcommand of the instance manager

Except for postgres and edb_audit that have the aforementioned structures, all other possible values just have msg set to the escaped message that's logged.