Open cosmo0920 opened 7 years ago
Yes, this PR does not break compatibility for Fluentd v0.14.
Note that chunk compatibility will be broken with this change. If users use this plugin with this patch, he/she should ensure that there is no unprocessed buffer. This breaking changes is operational issue same as migrating to use v0.14 from v0.12 series of Fluentd.
Hmm, I'll notify again @warmfusion
Do both end's of the plugin need to be modified so the tag data in the metadata is recognised; ie should in_amqp also be modified?
Can the README be updated to account for the new configuration option and its behaviour?
Should the chunk_keys
configuration even be exposed to end-users through configuration? Why not just use the tag
value as currently defined? I assume you'd want to be able to set multiple chunk_key values (otherise the plural name is misleading) - but I see no mechanism for such a thing to be defined?
Should the implementation silently use the tag value as currently defined, or is there something fundamental i'm missing here?
Do both end's of the plugin need to be modified so the tag data in the metadata is recognised; ie should in_amqp also be modified?
No we don't need to this. Input plugins do not have chunk -- i.e. this optimization is effectively working with buffered output plugins, which handle chunk.
I think that this PR is too hard to understand metadata mechanism. It is early to add this reducing feature in 3rd party plugin.
I leave this PR as-is for now.
In v0.14,
tag
can be included in chunk's metadata. It makes to reduce chunk size.