[searchisko-dev] Access to the elasticsearch service itself?

[Lukas Vlcek]

Hi Pete,

In general, Searchisko provides only HTTP REST interface now thus I doubt that existing Ruby|Python|Perl|whatever elasticsearch clients can be used to send valid requests to Searchisko instances (like to our DCP).

The API that is available is quite limited compared to what raw Elasticsearch provides. But it would be too risky to open the underlying Elasticsearch API to our clients.

First, Elasticsearch does not employ any security/authorization mechanisms (and IMO it will hardly have it in the near future). So once you get access to Elasticsearch API you can do anything (including deleting all data, running infinite scripts, shutting down whole cluster, ...).

Second, API in Elasticsearch is still subject to changes (still this is happening, and for example in upcoming 1.x there is plan for completely new faceting API).

In this sense you can think of Searchisko as a proxy to Elasticsearch centred around simplification of common operations for building a search for "typical community" content. On top of that Searchisko helps/automate some content normalization operations for which it is using "internal" indices. Such indices should not be manipulated directly by clients via Elasticsearch API (it is like internal database tables).

However, we might add a possibility to send a raw requests to underlying Elasticsearch server in the future but this needs to be done very carefully.

Regards,
Lukas Vlcek
jboss.org Development Team

----- Original Message -----
> Adding the list anyway :-)
> 
> On 15 Aug 2013, at 19:31, Jason Porter <jporter at redhat.com> wrote:
> 
> > Hi all, I'd originally sent this to Lukas, but if any of you have
> > suggestions, I'm certainly open to them. Also, are we all on the
> > searchisko mailing list?
> > 
> > ----- Forwarded Message -----
> > From: "Jason Porter" <jporter at redhat.com>
> > To: "Lukas Vlcek" <lvlcek at redhat.com>
> > Cc: "Pete Muir" <pmuir at redhat.com>
> > Sent: Wednesday, August 14, 2013 2:57:47 PM
> > Subject: Access to the elasticsearch service itself?
> > 
> > While playing around with the Search API for the DCP, I came across some
> > ruby (and other langs) clients for elasticsearch itself. This actually
> > started when I was trying to get a list of contributors, maybe I'm going
> > about this wrong but it doesn't seem to be giving me what I'm looking for:
> > 
> > # Contributor search
> > response = conn.get do |req|
> >  req.url '/v1/rest/search',
> >          {
> >              :dcp_type => 'contributor',
> >              #:type => 'sys_contributors',
> >              :query => 'Proctor'
> >          }
> > end
> > 
> > If I send the 'type' param, then I get an exception back from the server
> > (I'm using https://dcp-jbossorgdev.rhcloud.com/). I'm receiving 1513
> > results back from the server, I would expect to receive one, maybe as many
> > as 10, but I don't think we have many contributors with a name of Proctor.
> > If we are able to use one of the other elasticsearch clients (not for
> > indexing, just searching) something like this may be easier.
> 
> 
>