Here's a couple of JIRAs that are targeting Hadoop 2.6, which may be of relevance
https://issues.apache.org/jira/browse/HADOOP-11064: UnsatisifedLinkError with hadoop 2.4 JARs on hadoop-2.6 due NativeCRC32 method changes Changes in the native API to deal with CRCs mean if you upload the hadoop-2.4. jars to a 2.6-SNAPSHOT YARN cluster today, you get a stack trace rather than an AM. I think the way Twill works means that'll be the default outcome. The current patch will just stop native lib loading, which will kill performance instead. I would recommend expressing opinions on the matter. https://issues.apache.org/jira/browse/YARN-913: Add a way to register long-lived services in a YARN cluster. Twill already has its own registry; this proposes a standard one which the RM can set up securely, and even clean up when apps & containers go away -even if the entries aren't marked as ephemeral. (you need better persistence if you want an entry to outlive an attempt restart). I believe twill will be able to bind to this if you want to —for now comments are encouraged to make sure that it meets your needs without it trying to be the "one registry to bind them" epic bit of infrastructure. For now it's ZK based; I'm planning a REST API later, with long-haul support too. Later == hadoop 2.7+ -Steve -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
