manifoldcf-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Karl Wright <daddy...@gmail.com>
Subject Re: SolrCloud connection
Date Wed, 05 Nov 2014 14:01:51 GMT
Hi Luca,

While I'm not going to be able to help you much with details about how Solr
Cloud differs from a single stand-alone ManifoldCF, I can answer certain
questions easily:

bq. First of all, we note that once you have selected a "Solr type" option
(Single server, SolrCloud), only one of the two following tabs (Server,
Zookeeper) should be available, not both, as it actually is.

MCF does not add or suppress tabs for any connector based on what you
select in other tabs.  So this behavior is working as designed.

bq. Secondly, and more important, using the Zookeeper tab we couldn't
connect with our Solr Cloud cluster.  By the way, after giving the correct
"Zookeeper host" and "Collection name", we couldn't find a valid "Znode
path". We tried many options but no one is working.

It would be great to get a screen shot of the view page for your SolrCloud
connection.  Nobody else has complained about being unable to connect
properly to Zookeeper.  The connector's SolrCloud zookeeper instance info
is quite generic and should work fine in a broad range of situations,
unless somebody in the Solr world has changed how Zookeeper is managed in
SolrJ.  What version of Solr are you using?

bq. As an alternative to the Solr Cloud connection, we tried a Single
Server connection...

That's a good way to debug, although I have no idea if going directly to a
load balancing node would work.  That's not how SolrJ expects to be
indexing via SolrCloud. It would be better to set up a single stand-alone
instance as an example and see if you can get that to work first.


I'm going to have to defer to folks like Alessandro to give further hints
about your Solr connector configuration at this point.

Thanks,
Karl


On Wed, Nov 5, 2014 at 8:48 AM, Basso Luca <LBasso@regione.emilia-romagna.it
> wrote:

>  Hi Karl,
>
> we're using a Web Repository Connection and a Solr Output Connection in
> MCF 1.7.1 .
>
> Specifically we need to address a Solr Cloud cluster.
>
> In the official Documentation we read the following:
>
>
>
> "When you create a Solr output connection, multiple configuration tabs
> appear. The first tab is the 'Solr type' tab.
>
> Here you select whether you want your connection to communicate to a
> standalone Solr instance, or to a Solr Cloud cluster.
>
> Select which kind of Solr installation you want to communicate with. Based
> on your selection, you can proceed to either the 'Server' tab
>
> (if a standalone instance) or to the 'ZooKeeper' tab (if a Solr Cloud
> cluster).
>
> The 'Server' tab allows you to configure the HTTP parameters appropriate
> for communicating with a standalone Solr instance.
>
> The 'Zookeeper' tab allows your to configure the connection type to
> communicate with a Solr Cloud cluster."
>
>
>
> First of all, we note that once you have selected a "Solr type" option
> (Single server, SolrCloud), only one of the two following tabs
>
> (Server, Zookeeper) should be available, not both, as it actually is.
>
> Secondly, and more important, using the Zookeeper tab we couldn't connect
> with our Solr Cloud cluster.
>
> By the way, after giving the correct "Zookeeper host" and "Collection
> name", we couldn't find a valid "Znode path".
>
> We tried many options but no one is working.
>
> Please see also our Solr Cloud tree in the underlying screenshot.
>
>
>
> As an alternative to the Solr Cloud connection, we tried a Single Server
> connection, directly addressing the Load Balancer
>
> which is in front of our Solr cloud servers.
>
> Everything seemed to be ok on the MCF side, but actually no one document
> was indexed in Solr.
>
> That looks a little bit surprising, because a curl command like the
> following is working indeed:
>
> curl "
> http://10.10.80.185:80/solr/myCollection/update/extract?literal.id=1&fmap.content=body&commit=true"
> -F "myfile=@myDocument"
>
> Lastly, we note that when we abort our Single Server job, it never reaches
> the "Done" status, but rather stops in the "End notification" status.
>
>
>
> Can you give us any suggestions regarding the two above mentioned
> scenarios?
>
> Thanks.
>
>
>
> Luca
>
>
>
>

Mime
View raw message