Re: SolrCloud: Different replicationFactor for different shards in same collection

[Nawab Zada Asad Iqbal <kh...@gmail.com>]

Thanks Erick


This is for future. I am exploring to use a custom sharding scheme (which
will require modification in Solr code) together with the benefits of
SolrCloud.



Thanks
Nawab



On Tue, Jul 31, 2018 at 4:51 PM, Erick Erickson <er...@gmail.com>
wrote:

> Sure, just use the Collections API ADDREPLICA command to add as many
> replicas for specific shards as you want. There's no way to specify
> that at creation time though.
>
> Some of the new autoscaling can do this automatically I believe.
>
> I have to ask what it is about your collection that this is true. If
> you're using the default composite id routing having one shard get
> substantially more queries than the others is unexpected.
>
> If you're using implicit routing then it's perfectly understandable.
>
> Best,
> Erick
>
> On Tue, Jul 31, 2018 at 4:12 PM, Nawab Zada Asad Iqbal <kh...@gmail.com>
> wrote:
> > Hi,
> >
> > I am looking at Solr 7.x and couldn't find an answer in the
> documentation.
> > Is it possibly to specify different replicationFactor for different
> shards
> > in same collection? E.g. if a certain shard is receiving more queries
> than
> > rest of the collection  I would like to add more replicas for it to help
> > with the query load.
> >
> >
> >
> > Thanks
> > Nawab
>

Re: SolrCloud: Different replicationFactor for different shards in same collection

[Erick Erickson <er...@gmail.com>]

This feels like more work than necessary, especially the bit:
"which will require modification in Solr code".

If your needs are to co-locate various groups of documents
on specific nodes, composite id (the default) routing has
the ability to cluster docs together, see:
https://lucene.apache.org/solr/guide/6_6/shards-and-indexing-data-in-solrcloud.html
the "document routing" section. You can also route queries to
those shards only, see:
https://lucidworks.com/2013/06/13/solr-cloud-document-routing/

If that isn't sufficient, using "implicit" routing allows you to
send document to specific shards.

True, in both cases the _client_ has to assign the doc to a particular
shard based on whatever criteria you need, but that seems like less
work than changing Solr code.

Best,
Erick

On Tue, Jul 31, 2018 at 5:20 PM, Nawab Zada Asad Iqbal <kh...@gmail.com> wrote:
> Thanks Erick
>
>
> This is for future. I am exploring to use a custom sharding scheme (which
> will require modification in Solr code) together with the benefits of
> SolrCloud.
>
>
>
> Thanks
> Nawab
>
>
>
> On Tue, Jul 31, 2018 at 4:51 PM, Erick Erickson <er...@gmail.com>
> wrote:
>
>> Sure, just use the Collections API ADDREPLICA command to add as many
>> replicas for specific shards as you want. There's no way to specify
>> that at creation time though.
>>
>> Some of the new autoscaling can do this automatically I believe.
>>
>> I have to ask what it is about your collection that this is true. If
>> you're using the default composite id routing having one shard get
>> substantially more queries than the others is unexpected.
>>
>> If you're using implicit routing then it's perfectly understandable.
>>
>> Best,
>> Erick
>>
>> On Tue, Jul 31, 2018 at 4:12 PM, Nawab Zada Asad Iqbal <kh...@gmail.com>
>> wrote:
>> > Hi,
>> >
>> > I am looking at Solr 7.x and couldn't find an answer in the
>> documentation.
>> > Is it possibly to specify different replicationFactor for different
>> shards
>> > in same collection? E.g. if a certain shard is receiving more queries
>> than
>> > rest of the collection  I would like to add more replicas for it to help
>> > with the query load.
>> >
>> >
>> >
>> > Thanks
>> > Nawab
>>