Adds buffer_ignore_exceeded_chunk option to Buffer#306
Conversation
Controls whether exceeded_chunk should raise an exception (and break the chain) or log and ignore the exception.
|
This code will enqueue at lib/fluent/buffer.rb#L207-L209 even if queue size ecceeds |
(even if using @ignore_exceeded_chunk)
|
@sonots added a test case |
|
-1 for this change. Here're two reasons: First one is a philosophical thing. Fluentd's importance is 'simplicity'. The more we have options, the more beginners need to learn. Second, you still have a limit on disk volume anyway. And if we have a disk full error while writing to the buffer, we possibly couldn't recover the data. What we want is an application level handling here. Is just having a larger value for |
|
@kzk yeah, for individual cases, increasing |
Controls whether exceeded chunk should raise an exception (and break the chain) or log and ignore
the exception. As discussed in #296