fluentd log driver: default retry wait is passed in nanoseconds #5320

In pkg/logging/fluentd_logger.go the default is defaultRetryWait = 1000 * time.Millisecond and retryWait := int(defaultRetryWait), which is 1000000000. fluent-logger-golang reads Config.RetryWait as milliseconds, and the code already converts a user-set fluentd-retry-wait with temp.Milliseconds().

Without fluentd-retry-wait the logger therefore gets about 11 days of retry wait. Its backoff caps the wait at MaxRetryWait (60000 ms), so a reconnect waits a minute instead of the intended one second.

Using defaultRetryWait.Milliseconds() fixes it. The existing tests compared against int(defaultRetryWait), so they matched the bug. I have a patch that changes them to check the value in milliseconds.