https://github.com/postgrespro/pg_logging
PostgreSQL logging interface
https://github.com/postgrespro/pg_logging
logging monitoring postgresql
Last synced: 6 days ago
JSON representation
PostgreSQL logging interface
- Host: GitHub
- URL: https://github.com/postgrespro/pg_logging
- Owner: postgrespro
- License: other
- Created: 2018-06-09T10:29:36.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2018-12-27T10:21:42.000Z (over 6 years ago)
- Last Synced: 2025-04-24T10:48:42.738Z (7 days ago)
- Topics: logging, monitoring, postgresql
- Language: C
- Homepage:
- Size: 56.6 KB
- Stars: 15
- Watchers: 32
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
[](https://travis-ci.org/postgrespro/pg_logging)
[](https://raw.githubusercontent.com/postgrespro/pg_logging/master/LICENSE)
pg_logging
=================
PostgreSQL logging interface.
Installation
-------------
# make sure that directory with pg_config in PATH or specify it with PG_CONFIG variable
make install
# in postgresql.conf add:
shared_preload_libraries = 'pg_logging'
# install the extension
> CREATE EXTENSION pg_logging;
Available functions
--------------------
get_log(
flush bool default true
)
get_log(
from_position int
)
This function is used to fetch the logged information. The information is
similar to the data that postgres writes to log files.
`flush` means that fetched data will not be returned on next query calls. By
default it's true.
`from_position` is used as fail-safe case, when a client specifies until
which position it already has data. This position should be equal to
`position` field from `log_item`. If there was wraparound there is a chance
that the position will be invalid and query will raise an error. In this case
the client should use `get_log(flush bool)` function (and possibly increase
the ring buffer size).
Logs are stored in the ring buffer which means that non fetched data will
be rewritten in the buffer wraparounds. Since reading position should be
accordingly moved on each rewrite it could slower down the database.
`get_log` function returns rows of `log_item` type. `log_item` is specified as:
create type log_item as (
log_time timestamp with time zone,
level int,
pid int,
line_num bigint, /* log line number */
appname text,
start_time timestamp with time zone, /* backend start time */
datid oid, /* database id */
errno int,
errcode int,
errstate text,
message text,
detail text,
detail_log text,
hint text,
context text,
context_domain text,
domain text,
internalpos int,
internalquery text,
userid oid,
remote_host text,
command_tag text,
vxid text, /* virtual transaction id */
txid bigint, /* transaction id */
query text,
query_pos int,
position int
);
`error_level` type
-------------------
The extension installs special type called `error_level` which can be used to
get textual representation of `level` field from `log_item`. To do that
just add something like `level::error_level` to the columns list in your query.
Options
---------
pg_logging.buffer_size (10240) - size of internal ring buffer in kilobytes.
pg_logging.enabled (on) - enables or disables the logging.
pg_logging.ignore_statements (off) - skip statements lines if `log_statement=all`
pg_logging.set_query_fields (on) - set query and query_pos fields.