You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Upayavira <uv...@odoko.co.uk> on 2011/06/10 00:26:30 UTC
SolrCloud questions
I'm exploring SolrCloud for a new project, and have some questions based
upon what I've found so far.
The setup I'm planning is going to have a number of multicore hosts,
with cores being moved between hosts, and potentially with cores merging
as they get older (cores are time based, so once today has passed, they
don't get updated).
First question: The solr/conf dir gets uploaded to Zookeeper when you
first start up, and using system properties you can specify a name to be
associated with those conf files. How do you handle it when you have a
multicore setup, and different configs for each core on your host?
Second question: Can you query collections when using multicore? On
single core, I can query:
http://localhost:8983/solr/collection1/select?q=blah
On a multicore system I can query:
http://localhost:8983/solr/core1/select?q=blah
but I cannot work out a URL to query collection1 when I have multiple
cores.
Third question: For replication, I'm assuming that replication in
SolrCloud is still managed in the same way as non-cloud Solr, that is as
ReplicationHandler config in solrconfig? In which case, I need a
different config setup for each slave, as each slave has a different
master (or can I delegate the decision as to which host/core is its
master to zookeeper?)
Thanks for any pointers.
Upayavira
---
Enterprise Search Consultant at Sourcesense UK,
Making Sense of Open Source
Re: SolrCloud questions
Posted by Mark Miller <ma...@gmail.com>.
All of Solr is still under development ;)
If you do a bit of searching, you might find slides/video of a talk about how http://www.loggly.com/ has been using SolrCloud in production for some time now. They have been pretty happy with it based on what they have said.
Still need to tackle the indexing side, and there is always more to do in terms of ease of use and additional features, but I think what we have done is fairly solid.
- Mark
On Jun 10, 2011, at 7:25 AM, Mohammad Shariq wrote:
> I am also planning to move to SolrCloud;
> since its still in under development, I am not sure about its behavior in
> Production.
> Please update us once you find it stable.
>
>
> On 10 June 2011 03:56, Upayavira <uv...@odoko.co.uk> wrote:
>
>> I'm exploring SolrCloud for a new project, and have some questions based
>> upon what I've found so far.
>>
>> The setup I'm planning is going to have a number of multicore hosts,
>> with cores being moved between hosts, and potentially with cores merging
>> as they get older (cores are time based, so once today has passed, they
>> don't get updated).
>>
>> First question: The solr/conf dir gets uploaded to Zookeeper when you
>> first start up, and using system properties you can specify a name to be
>> associated with those conf files. How do you handle it when you have a
>> multicore setup, and different configs for each core on your host?
>>
>> Second question: Can you query collections when using multicore? On
>> single core, I can query:
>>
>> http://localhost:8983/solr/collection1/select?q=blah
>>
>> On a multicore system I can query:
>>
>> http://localhost:8983/solr/core1/select?q=blah
>>
>> but I cannot work out a URL to query collection1 when I have multiple
>> cores.
>>
>> Third question: For replication, I'm assuming that replication in
>> SolrCloud is still managed in the same way as non-cloud Solr, that is as
>> ReplicationHandler config in solrconfig? In which case, I need a
>> different config setup for each slave, as each slave has a different
>> master (or can I delegate the decision as to which host/core is its
>> master to zookeeper?)
>>
>> Thanks for any pointers.
>>
>> Upayavira
>> ---
>> Enterprise Search Consultant at Sourcesense UK,
>> Making Sense of Open Source
>>
>>
>
>
> --
> Thanks and Regards
> Mohammad Shariq
- Mark Miller
lucidimagination.com
BERLIN BUZZWORDS JUNE 6-7TH, 2011
Re: SolrCloud questions
Posted by Upayavira <uv...@odoko.co.uk>.
Mohammad,
There are two sides to using SolrCloud in production - the SolrCloud
code, and the Solr 4.0 code that it is a part of.
You can reduce the risk of being caught out by Solr/Lucene 4.0 changes
(e.g. index structure changes) by using a Lucene 3.0 index format within
Solr 4.0. While there's still risk involved in using an unreleased
product, you'll have increased your chances of stability.
Still hoping someone has answers to my original questions...
Upayavira
On Fri, 10 Jun 2011 10:55 +0530, "Mohammad Shariq"
<sh...@gmail.com> wrote:
> I am also planning to move to SolrCloud;
> since its still in under development, I am not sure about its behavior in
> Production.
> Please update us once you find it stable.
>
>
> On 10 June 2011 03:56, Upayavira <uv...@odoko.co.uk> wrote:
>
> > I'm exploring SolrCloud for a new project, and have some questions based
> > upon what I've found so far.
> >
> > The setup I'm planning is going to have a number of multicore hosts,
> > with cores being moved between hosts, and potentially with cores merging
> > as they get older (cores are time based, so once today has passed, they
> > don't get updated).
> >
> > First question: The solr/conf dir gets uploaded to Zookeeper when you
> > first start up, and using system properties you can specify a name to be
> > associated with those conf files. How do you handle it when you have a
> > multicore setup, and different configs for each core on your host?
> >
> > Second question: Can you query collections when using multicore? On
> > single core, I can query:
> >
> > http://localhost:8983/solr/collection1/select?q=blah
> >
> > On a multicore system I can query:
> >
> > http://localhost:8983/solr/core1/select?q=blah
> >
> > but I cannot work out a URL to query collection1 when I have multiple
> > cores.
> >
> > Third question: For replication, I'm assuming that replication in
> > SolrCloud is still managed in the same way as non-cloud Solr, that is as
> > ReplicationHandler config in solrconfig? In which case, I need a
> > different config setup for each slave, as each slave has a different
> > master (or can I delegate the decision as to which host/core is its
> > master to zookeeper?)
> >
> > Thanks for any pointers.
> >
> > Upayavira
> > ---
> > Enterprise Search Consultant at Sourcesense UK,
> > Making Sense of Open Source
> >
> >
>
>
> --
> Thanks and Regards
> Mohammad Shariq
>
---
Enterprise Search Consultant at Sourcesense UK,
Making Sense of Open Source
Re: SolrCloud questions
Posted by Mohammad Shariq <sh...@gmail.com>.
I am also planning to move to SolrCloud;
since its still in under development, I am not sure about its behavior in
Production.
Please update us once you find it stable.
On 10 June 2011 03:56, Upayavira <uv...@odoko.co.uk> wrote:
> I'm exploring SolrCloud for a new project, and have some questions based
> upon what I've found so far.
>
> The setup I'm planning is going to have a number of multicore hosts,
> with cores being moved between hosts, and potentially with cores merging
> as they get older (cores are time based, so once today has passed, they
> don't get updated).
>
> First question: The solr/conf dir gets uploaded to Zookeeper when you
> first start up, and using system properties you can specify a name to be
> associated with those conf files. How do you handle it when you have a
> multicore setup, and different configs for each core on your host?
>
> Second question: Can you query collections when using multicore? On
> single core, I can query:
>
> http://localhost:8983/solr/collection1/select?q=blah
>
> On a multicore system I can query:
>
> http://localhost:8983/solr/core1/select?q=blah
>
> but I cannot work out a URL to query collection1 when I have multiple
> cores.
>
> Third question: For replication, I'm assuming that replication in
> SolrCloud is still managed in the same way as non-cloud Solr, that is as
> ReplicationHandler config in solrconfig? In which case, I need a
> different config setup for each slave, as each slave has a different
> master (or can I delegate the decision as to which host/core is its
> master to zookeeper?)
>
> Thanks for any pointers.
>
> Upayavira
> ---
> Enterprise Search Consultant at Sourcesense UK,
> Making Sense of Open Source
>
>
--
Thanks and Regards
Mohammad Shariq
Re: SolrCloud questions
Posted by Mark Miller <ma...@gmail.com>.
On Jun 10, 2011, at 12:26 AM, Upayavira wrote:
> I'm exploring SolrCloud for a new project, and have some questions based
> upon what I've found so far.
>
> The setup I'm planning is going to have a number of multicore hosts,
> with cores being moved between hosts, and potentially with cores merging
> as they get older (cores are time based, so once today has passed, they
> don't get updated).
>
> First question: The solr/conf dir gets uploaded to Zookeeper when you
> first start up, and using system properties you can specify a name to be
> associated with those conf files. How do you handle it when you have a
> multicore setup, and different configs for each core on your host?
Upload each set of configs with a different name, and for each core tell it what 'set' of configs to use.
>
> Second question: Can you query collections when using multicore? On
> single core, I can query:
>
> http://localhost:8983/solr/collection1/select?q=blah
>
> On a multicore system I can query:
>
> http://localhost:8983/solr/core1/select?q=blah
>
> but I cannot work out a URL to query collection1 when I have multiple
> cores.
see http://wiki.apache.org/solr/SolrCloud#Distributed_Requests
I think someone has implemented a couple of those that are stated as not done - will have to check JIRA.
I really need to get back into SolrCloud work - it's been too long...
>
> Third question: For replication, I'm assuming that replication in
> SolrCloud is still managed in the same way as non-cloud Solr, that is as
> ReplicationHandler config in solrconfig?
Correct.
> In which case, I need a
> different config setup for each slave, as each slave has a different
> master
System properties and other standard ways of doing master/slave in one config should still work?
> (or can I delegate the decision as to which host/core is its
> master to zookeeper?)
I once started down this path, but never finished it. I still consider it somewhat low hanging fruit - but when we do the true SolrCloud indexing side, replication won't likely be part of it. Of course, when will that happen...so perhaps we can do something for this soon rather than later? Easy to make lots of plans while sick and in bed though.
>
> Thanks for any pointers.
>
> Upayavira
> ---
> Enterprise Search Consultant at Sourcesense UK,
> Making Sense of Open Source
>
- Mark Miller
lucidimagination.com