Keith Gregory

@yagrxu - are you saying that you see the timeout message but then see messages written to CloudWatch? This should not be possible: the normal cause of a timeout during...

After looking at the code, I don't think that wait is required for either Logback or Log4J2.x, so am going to remove it. I'll probably use a configurable startup timeout...

Update: my current estimate is early December.

@yagrxu - if you're willing to build it yourself, I've merged this work into the trunk. It passes all integration tests except for synchonous mode (which hopefully you're not using)....

This is now on Maven Central.

@anoopnarang Thanks for the well-documented request. It seems like a reasonable addition. My one concern is that `ILoggingEvent.getKeyValuePairs()` was added in 1.3, so support for it would break backwards compatibility...

Looking into this, and 1.3 also made a breaking change to the `AccessEvent` constructor. I'm going to do a first pass using reflection, then see whether it's even worthwhile to...

Thanks for letting me know, but I do think it's a good enhancement. If you're interested, the first pass, using reflection, is in branch `dev-3.2.1`, but is missing tests and...

Ran a [benchmark](https://github.com/kdgregory/log4j-aws-appenders/files/15141882/KeyValueExtractionBenchmark.txt) to compare reflection with direct access via the new classes. Direct access was 996 ms for 10MM reps, while reflection took 1153 ms. I don't see any...

Everything's done, but integration tests are failing due to what appears to be an AWS bug: sending a batch of records immediately after creating the stream fails with `ResourceNotFoundException`. Subsequent...