solcloud on production

[Midas A <te...@gmail.com>]

Hi All ,

we are moving from master slave architecture to solr cloud architecture .
so i would like to know following

-  what kind of challenges we can face on production  .

-  Is there any drawback of solrcloud

-  How solr cloud distributes requests between nodes and how node will
behave on heavy traffic .

- Is there any way to shard node with custom logic




Regards,
MA

Re: solcloud on production

["Kelly, Frank" <fr...@here.com>]

I have found that ZooKeeper is the weak link for SolrCloud especially if
deployed on a Cloud environment (AWS).

https://issues.apache.org/jira/browse/SOLR-3274

https://issues.apache.org/jira/browse/SOLR-8868


And others see 
https://issues.apache.org/jira/browse/ZOOKEEPER-2112?jql=project%20%3D%20ZO
OKEEPER%20AND%20issuetype%20%3D%20Bug%20AND%20status%20in%20(Open%2C%20%22I
n%20Progress%22%2C%20Reopened)%20AND%20text%20~%20%22quorum%22


The kicker is that is ZK loses and quorum but gets it back we not only
have to restart Solr servers but also those services who are clients :-(

-Frank

Frank Kelly
Principal Software Engineer
Predictive Analytics Team (SCBE/HAC/CDA)

HERE 
5 Wayside Rd, Burlington, MA 01803, USA
42° 29' 7" N 71° 11' 32² W

 <http://360.here.com/>   <https://twitter.com/here>
<https://www.facebook.com/here>    <https://linkedin.com/company/heremaps>
   <https://www.instagram.com/here>







On 4/8/16, 2:39 AM, "Midas A" <te...@gmail.com> wrote:

>Hi All ,
>
>we are moving from master slave architecture to solr cloud architecture .
>so i would like to know following
>
>-  what kind of challenges we can face on production  .
>
>-  Is there any drawback of solrcloud
>
>-  How solr cloud distributes requests between nodes and how node will
>behave on heavy traffic .
>
>- Is there any way to shard node with custom logic
>
>
>
>
>Regards,
>MA

Re: solcloud on production

[Shawn Heisey <ap...@elyograg.org>]

On 4/8/2016 12:39 AM, Midas A wrote:
> we are moving from master slave architecture to solr cloud architecture .
> so i would like to know following
>
> -  what kind of challenges we can face on production  .

This is extremely difficult to answer.  It would depend on the skill
level of the admin who does the deployment, and how well they understand
SolrCloud.  You have this mailing list and the IRC channel as resources
to help you through any problems.

Frank's response mentions zookeeper as a weak link ... I'm not sure this
is true.  If there are problems with your networking or your server
stability, then zookeeper may have issues, but these are not problems
with zookeeper itself.

> -  Is there any drawback of solrcloud

Depends on whether you consider "three servers minimum" (for zookeeper)
for a fault-tolerant install to be a drawback or not.

SolrCloud is a slightly more difficult to set up, but tends to be easier
to maintain after install.

> -  How solr cloud distributes requests between nodes and how node will
> behave on heavy traffic .

For indexing, cloud forwards new documents to the leader of the correct
shard.  The leader of the shard will index the documents locally and
also send those documents to all other replicas for that shard, which
will each index them independently.

For queries, cloud will randomly pick replicas in the cloud to serve
requests.  If there are replicas for the requested collection on the
local machine, the request *might* be served locally rather than
traverse the network again.

> - Is there any way to shard node with custom logic

You could probably write your own custom Router class and change the
router on the collection to use it, although I do not know how to do
this.  Once somebody finds out, a page on the old wiki, with a code
example, would be a good idea.  I don't know if that kind of
documentation belongs in the reference guide.

Thanks,
Shawn