David,
On each machine on which either ES or a client is deployed, we have the
following directory which contains all of the jars that are packaged with
ES:
/opt/db/current/elasticsearch-1.3.4/lib
Then the java command's -classpath includes
/opt/db/current/elasticsearch-1.3.4/lib/* (along with our own custom jars
via /opt/db/lib/*) and everything works fine.
As for additional 3rd party jars, I have the following:
- Jackson. The full library is used instead of the one inside ES.
- Netty. This was needed for my own REST API which hides ES and contains
the business logic. I couldn't figure out how to easily use the shaded
version inside ES, and the "real" Netty is as easy to use as falling off a
log.
- The LMAX Disruptor .jar file. This thing combines nicely with Netty and
wow! Netty and application thread counts remain low even under heavy loads.
Everything else I get directly from ES. And I love the way it shades its
versions of Netty and Jackson so it's very easy for my own app to cherry
pick what it wants from ES and what it prefers outside of ES.
We could use maven, I suppose, but we don't. Instead, we package all of the
jars into a zip archive after our application is built against a specific
ES version. And then that single self-contained zip archive is installed
where it is needed. And there is no need for an external or internal maven
repo. Not a big deal for us.
All in all, it's much like how Elasticsearch itself is packaged and
distributed: A zip archive that I download from the web site. I would never
use a .deb or .rpm since the version that I want is always on the web site.
And I believe there is a maven repo but the .zip archive links are right on
the web site, and we don't update all that often (regularly, but I don't
thrash our deployment folks).
It sounds complicated, I suppose. But that was only once, and it's been
easy to manage and develop against, easy to deploy, and makes me look very,
very good to our deployment folks.
Brian
P.S. I don't use Guice or Spring. I don't see any problem with the new
operator, and the services I create are fast, rock-solid, easy to configure
and deploy, and that puts me light-years ahead of much of the pack. But
this is another topic altogether! 
On Saturday, November 15, 2014 12:24:28 AM UTC-5, David Pilato wrote:
Hi Brian,
I think I'm missing something.
At the end you still have the full elasticsearch jars, right?
What is the difference with having that as a maven dependency?
Is it a way for not getting all elasticsearch dependencies which are
shaded in elasticsearch jar such as Jackson, Guice,... ?
David
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/01454216-da84-49c0-85e6-62efe4ad535d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.