Checked Java DNS resolution caching: ``` sun.net.InetAddressCachePolicy.get(); sun.net.InetAddressCachePolicy.getNegative(); ```
and those return 30 and 10 respectively. So it seems fine and it shouldn't cache for too long. On 2021/05/14 12:55:00, Michał Łowicki <mlowi...@gmail.com> wrote: > Hey, > > Had incident where one broker died and got later different IP address. Some > clients / pods (everything lives on K8s) detected IP change and logged: > > [Producer clientId=producer-1] Hostname for node 10000 changed from XXXX to > > YYYY. > > > (logged by org.apache.kafka.clients.ClusterConnectionStates) but many > didn't and they couldn't actually send anything, continuously timing out. > Any clue if there were some know issues with that and maybe something has > been fixed in > 2.4.1? > > In the meantime I'm also checking DNS caching but nothing at OS-level but > still verifying JVM (I'm using Java 14). > > -- > BR, > Michał Łowicki >
