You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In #87939 we merged support for request/response logging to the KP. One of the issues that surfaced was that logs suddenly became very cluttered as the LogMeta that's attached to the response logs can be quite large.
The temporary workaround we discussed was to simply remove meta from the default pattern layout, so that folks would need to "opt-in" to view the meta when using pattern layout by defining a custom appender.
However, a more ideal approach might be to have a mechanism where you could specify context-specific appender defaults to apply when no other configuration is provided, as discussed in #90457. This might be a little similar to the Elasticsearch default logging config file.
A solution like this could solve a variety of problems, such as:
specific loggers that output too much data (e.g. for http response logs, default to a custom appender that has no meta)
loggers that benefit from a more specific pattern layout which differs from the default
migration scenarios where you might want to preserve a different BWC format for a particular logger (this one is hypothetical, we admittedly don't have a concrete need for this today)
We'd need to discuss how to implement this, but in general I'd imagine a dedicated file where the defaults are specified, and then get merged into the logging config's configSchema.defaultValue & applied whenever the logging system is upgraded.
In #87939 we merged support for request/response logging to the KP. One of the issues that surfaced was that logs suddenly became very cluttered as the
LogMetathat's attached to the response logs can be quite large.The temporary workaround we discussed was to simply remove
metafrom the defaultpatternlayout, so that folks would need to "opt-in" to view the meta when using pattern layout by defining a custom appender.However, a more ideal approach might be to have a mechanism where you could specify context-specific appender defaults to apply when no other configuration is provided, as discussed in #90457. This might be a little similar to the Elasticsearch default logging config file.
A solution like this could solve a variety of problems, such as:
Once we remove the injected "default" legacy appender in 8.0, it might be a good time to consider this as an enhancement.
We'd need to discuss how to implement this, but in general I'd imagine a dedicated file where the defaults are specified, and then get merged into the logging config's
configSchema.defaultValue& applied whenever the logging system is upgraded.