Description:
Greetings,
It looks like there is an issue or bug with the metrics provider. According to the documentation (https://dev.mysql.com/doc/mysql-operator/en/mysql-operator-properties.html#mysql-operator-...), we should be able to pass the exporter command-line arguments as options (InnoDBCluster.spec.metrics.options). I attempted this with the prom/mysqld_exporter image, including the required command-line arguments as options, the metrics: section form the manifest is furnished below.
metrics:
enable: true
image: prom/mysqld-exporter
monitor: true
monitorSpec:
labels:
release: prometheus-operator
options:
- --mysqld.address=localhost:3306
- --mysqld.username=metrics
However, it just ignores the operations, resulting in the metrics sidecar crash-looping with the below messages.
ts=2023-11-23T10:07:05.058Z caller=config.go:150 level=error msg="failed to validate config" section=client err="no user specified in section or parent"
ts=2023-11-23T10:07:05.058Z caller=mysqld_exporter.go:225 level=info msg="Error parsing host config" file=.my.cnf err="no configuration found"
It is because the command line arguments specified in the options: are not being recognized by the sidecar pod, resulting in the sidecar just running the mysql-process as shown below.
k run test --image prom/mysqld-exporter --command sleep infinity
pod/test created
k exec test -- mysqld_exporter
ts=2023-11-23T11:34:22.610Z caller=config.go:150 level=error msg="failed to validate config" section=client err="no user specified in section or parent"
ts=2023-11-23T11:34:22.610Z caller=mysqld_exporter.go:225 level=info msg="Error parsing host config" file=.my.cnf err="no configuration found"
The labels specified in the monitorSpec are also being ignored, resulting in the creation of a ServiceMonitor object without the mentioned labels.
monitorSpec:
labels:
release: prometheus-operator
How to repeat:
Enable metrics with the following parameters on an InnoDB cluster resource.
metrics:
enable: true
image: prom/mysqld-exporter
monitor: true
monitorSpec:
labels:
release: prometheus-operator
options:
- --mysqld.address=localhost:3306
- --mysqld.username=metrics
Once applied, the InnoDB pods will enter a crash loop due to issues with the metrics sidecar container. The metrics sidecar will display the following logs:
ts=2023-11-23T10:07:05.058Z caller=config.go:150 level=error msg="failed to validate config" section=client err="no user specified in section or parent"
ts=2023-11-23T10:07:05.058Z caller=mysqld_exporter.go:225 level=info msg="Error parsing host config" file=.my.cnf err="no configuration found"
Suggested fix:
