Hi. We use async net.logstash.logback.appender.LogstashTcpSocketAppender. It puts messages in buffer and sends them to Logstash occassionally. So if there are connection issues it will accumulate messages, which is expected. But we discovered that messages are huge (33 Mbs each). So we found all this data in ch.qos.logback.classic.spi.LoggingEvent.argumentArray field. We use Cassandra client called Hector, and if there is a connection issue it logs error passing a connection client as an argument. So GC cannot release this client because LoggingEvent has reference. And at this moment we already have formattedMessage, so there is no need to have argumentArray filled anymore as I understand. So the question is: why argumentArray is not being cleared after formatting the message? Is it by design? Please let me know if you know the answer! Thanks in advance! Andrey -- View this message in context: http://logback.10977.n7.nabble.com/Memory-leaks-if-messages-are-not-being-po... Sent from the Users mailing list archive at Nabble.com.