You will also want to set the replication on the directory/jars in HDFS to something high. I think in my testing I could not set it above 50.
-----Original Message----- From: Josh Elser [mailto:[email protected]] Sent: Thursday, February 20, 2014 12:30 PM To: [email protected] Subject: Re: best practice for deploying server side iterator jars A word of warning... https://issues.apache.org/jira/browse/VFS-487 Be careful about redeploying jars in "rapid" succession as it's possible that you'll get hit with the above issue. The classloader is currently configured to monitor the filesystem every second -- it might be prudent to make sure you wait at least 5s to make sure each file monitor sees the update. On 2/20/14, 5:53 AM, Brian Loss wrote: > Starting with Accumulo 1.5 there is a VFS class loader that is capable of loading jars from HDFS. If you have a look at the accumulo-site.xml in conf/examples/vfs-classloader, there is example configuration in the "VFS ClassLoader Settings" section. > > If you're on 1.4 or earlier, I don't believe there is any alternative to just making sure the jars are distributed to all tservers. > > On Feb 20, 2014, at 7:54 AM, Mike Hugo <[email protected]> wrote: > >> Hello, >> >> I'm wondering if there's an easy way to make sure a jar we have with custom server side iterators gets properly placed into the tserver classpath on each tserver (in ACCUMULO_HOME/lib/ext). We are about to integrate that into our devops scripts so that the jar is put in the right place as part of server install and configuration, but i was wondering - is there a way to put the jar into HDFS or something so that it's more "automagically" picked up by all the tservers? >> >> Thanks, >> >> Mike >
