My postgresql 9.3 logs are sprinkled with entries like these (redacted to protect the guilty):
STATEMENT: INSERT INTO table (key, value) VALUES ($1, $2)
ERROR: duplicate key value violates unique constraint "table_pkey"
DETAIL: Key (key)=(xyz) already exists.
STATEMENT: INSERT INTO table (key, value) VALUES ('xyz', '{...json...}')
I know the source of the problem (a race condition where two app servers do the same work) and that code gracefully handles the insertion failure, but for the life of me I can't seem to get PostgreSQL to stop logging the statement which generated the error to begin with.
I've set log_statements for this db to 'none' (which persists across connections):
show log_statement;
log_statement
---------------
none
(1 row)
But to no avail. These entries still keep hitting the logs. I wouldn't mind except the "values" from the logged statement are JSON blobs which aren't tiny. How can I keep PostgreSQL from logging whenever an insert violates the unique constraint?
log_error_verbosity
?log_min_messages
andlog_min_error_statement
: postgresql.org/docs/current/static/…log_statement
does not do the trick because it logs statements irregardless of errors. You want to modify what's logged in case of an error. Milen and a_horse provided the relevant settings for this.