Fluentd config incorrectly parses MariaDB logs
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
kolla-ansible |
Fix Released
|
Medium
|
Isaac Prior | ||
Rocky |
Fix Released
|
Medium
|
Mark Goddard | ||
Stein |
Fix Released
|
Medium
|
Mark Goddard | ||
Train |
Fix Released
|
Medium
|
Mark Goddard | ||
Ussuri |
Fix Released
|
Medium
|
Isaac Prior |
Bug Description
Fluentd parses mariadb.log in the MariaDB container when Monasca/Fluentd is enabled.
Two different log formats are used in mariadb.log (mysqld 10.1.20-MariaDB).
Log messages from the mysqld_safe process are parsed by fluentd correctly but others are not.
An example of the two log formats:
"""
190913 10:00:32 mysqld_safe Starting mysqld daemon with databases from /var/lib/mysql/
190913 10:00:32 mysqld_safe WSREP: Running position recovery with --log_error=
2019-09-13 10:00:33 140183735036096 [Note] /usr/libexec/mysqld (mysqld 10.1.20-MariaDB) starting as process 194 ...
190913 10:00:36 mysqld_safe WSREP: Recovered position 00000000-
2019-09-13 10:00:36 139854442735808 [Note] /usr/libexec/mysqld (mysqld 10.1.20-MariaDB) starting as process 229 ...
"""
As viewed in Kibana the '2019-09-13' format messages are interpretted as continuations of the '190913' format ones. The result is a multi-line log event that captures all messages until the next short-form time format.
The correct behaviour is to parse all messages that begin with a timestamp (of both types) as individual log messages.
Centos7 - 3.10.0-
Docker version 17.05.0-ce, build 89658be
Tested in Kolla-Ansible Vagrant environment - Stable/Rocky.
Kolla tag 8.0.0.0rc1
globals.yml includes:
"""
enable_monasca: "yes"
monasca_
"""
Changed in kolla-ansible: | |
importance: | Undecided → Medium |
Fix proposed to branch: master /review. opendev. org/686428
Review: https:/