Interesting project - first experiment

[Steve Blackmon]

Guys,

I think this is a great idea and I'd like to get a testbed running to
see how it holds up vs some of the other comparable in-memory and
relational backed solutions.

What I'm trying to do is deploy the war to a hosted tomcat; I have
tried cloudbees and elastic beanstalk.

Even though I've removed the web.xml reference to /etc/cassandra.yaml
and passed all of the parameters in via the context in web.xml, the
app still tries to load /etc/cassandra.yaml and bombs out from NPE.

Is this the expected behavior? How can I run the app without control
of the local filesystem?

Thanks

[Günter Ladwig]

Hi Steve,

this is expected, but maybe not desirable behavior ;) The only parameter you can set in web.xml is 'config-file', which is the location of cumulus.yaml (you do mean cumulus.yaml, right? cassandra.yaml is the Cassandra config file). All other parameters are taken from cumulus.yaml. Is there no location where your Tomcat has access to where you could put cumulus.yaml? If this is not possible I could change the code so that we try to get parameters from web.xml if cumulus.yaml is not found.

We are still working on a first release (before end of October hopefully), so please let us know if you have any others problems or suggestions!

Cheers,
Günter

[Steve Blackmon]

Thanks for the quick reply.

That would be a major improvement to usability IMO.  Since the intent is to make it easy to run this in a cloud environment, the servlet parameters should be able to provide all of the properties necessary to connect to cassandra and get started with the crud operations.

Steve Blackmon
Ravel
512-965-0451

[Günter Ladwig]

Hi,

I made the necessary changes, if the config-file parameter in web.xml is not specified or the config file is not found, all parameters are read from web.xml. You can update from SVN.

You need to specify cassandra-hosts, cassandra-keyspace, storage-layout, and mode (just like in cumulus.yaml).

A quick note on the mode, as this is not documented anywhere:

- In 'dataset' mode, the server returns RDF only for URIs under /resource (through a redirect to /data), similar to the dbpedia.org setup. This mode should be used when the CumulusRDF owns the namespace (i.e. CumulusRDF is hosted on the domain of the URIs).
In dataset mode you can also perform lookups on arbitrary triple patterns, for example like this:

http://localhost:8080/query?s=<http://dbpedia.org/resource/United_Kingdom>

(parameters s, p and o should of course be URL encoded, which I didn't do here)

- In 'proxy' mode, all requests are translated into lookups in the dataset. You can use this by specifying the CumulusRDF host as a proxy server. For example with curl:

curl -x localhost:8080 -H "Accept: application/rdf+xml" http://dbpedia.org/resource/United_Kingdom

This will return a description of http://dbpedia.org/resource/United_Kingdom (i.e., some triples where the URI appears as subject or object) from the data stored in CumulusRDF.

Cheers,
Günter