David Butler

Some more context in case you want to reproduce... We're using Amazon's Elasticache hosted Redis. When I run the following, it just hangs forever. ``` ruby require 'socket' socket=TCPSocket.open('yourserverhere.0001.use1.cache.amazonaws.com', 0)...

Hi, I imagine that you could implement a custom formatter used by the filtered log.

Hi, Tagged logging isn't something that exists in Ruby's standard library. It's available in ActiveSupport, and LogStashLogger has a compatible implementation. So naturally `Logger::Formatter`, which is in the Ruby standard...

Sure, that would be great. Are you planning to replace it with `kafka-ruby`?

Either way is fine. Thanks for helping out!

The UDP device also writes in batches because Ruby writes to sockets in batches. If you're seeing lots of duplicates, that means Ruby thought it couldn't write to the UDP...

Logging exceptions are written to STDERR by default. Are you checking there? `autoflush` doesn't disable batching. It flushes the buffer after each log message is written. Log messages can still...

That is indeed strange. It does appear that `Errno::EMSGSIZE` is not being correctly treated as unrecoverable. What version of LogStashLogger are you using? Dropping unrecoverable errors was released in 0.16.0.

I tried a small test, and got the result I expected: ```ruby logger = LogStashLogger.new(type: :udp, port: 5228) loop { 10.times { logger.info(Time.now); sleep(0.1) }; logger.info('1'*100_000); } E, [2017-05-13T14:55:40.690990 #16663]...

What version of Ruby are you using? In MRI 2.4.1, I get the error message `Message too long`, not `Message too large`