[newbie] questions about 3.6.0 and 4.x or 5.x ?

[Robert Hume <rh...@gmail.com>]

Hello, I'm hoping to get some quick advice from the Solr gurus out there ...



I’ve inherited a project that uses a Solr 3.6.0 deployment.   (Several
masters and several slaves – I think there are 6 Solr instances in total.)



I’ve been tasked with investigating if upgrading our 3.6.0 deployment will
improve performance – there’s a lot of data and things are getting slow,
apparently.



I’ve read Apache docs that from 3.6.x to 4.x there were improvements in
scalability and performance.



I see that from 4.x to 5.x that Solr is now a standable server and no
longer just a WAR running on Tomcat.





QUESTIONS:


A. Is it worth upgrading to 4.x or 5.x?  Will I see a big improvement in
performance?



B. Should I got to 4.x or 5.x?  Will 4.x be an easier upgrade path since
it's just a new WAR file?



C. In a nutshell ... what will the upgrade path look like, what kind of
steps am I in for, and how can I avoid trouble?




Any help is GREATLY appreciated!!


Rob

Re: [newbie] questions about 3.6.0 and 4.x or 5.x ?

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

To chime in, in certain cases the memory requirements for 4x (and 5x) are _much_
improved, see: https://lucidworks.com/blog/2012/04/06/memory-comparisons-between-solr-3x-and-trunk/

But as Shawn says, it's not a magic bullet.

Solr 5 requires Java 7, so that's one thing to be aware of. Plus, you
either have to upgrade
your indexes to 4.x, then install/upgrade to 5x or if you want to jump
straight from
3x to 5x, you need to re-index from scratch; Solr 5x will not read an
index created with
Solr 3x.

And rather than have a master/slave setup you'll probably want to
migrate to SolrCloud
as well, it's much easier to create/manage a cluster with shards with SolrCloud.

Best,
Erick



On Wed, Oct 21, 2015 at 12:28 PM, Shawn Heisey <ap...@elyograg.org> wrote:
> On 10/21/2015 12:41 PM, Robert Hume wrote:
>> I've inherited a project that uses a Solr 3.6.0 deployment.   (Several
>> masters and several slaves – I think there are 6 Solr instances in total.)
>>
>> I've been tasked with investigating if upgrading our 3.6.0 deployment will
>> improve performance – there’s a lot of data and things are getting slow,
>> apparently.
>>
>> I've read Apache docs that from 3.6.x to 4.x there were improvements in
>> scalability and performance.
>
> Performance does get better in newer versions, but for most use cases,
> there is NOT a night/day difference, just a minor speedup.  Upgrading
> *might* help, but even if it does, chances are that it will not
> completely solve the problem.
>
> The most common reason for Solr performance problems is that there is
> not enough memory.  That might mean the java heap is a little too small,
> but more frequently, it means that there's not enough memory in the
> server to cache the index contents effectively.
>
> General information:
>
> https://wiki.apache.org/solr/SolrPerformanceProblems
>
> Solr 3.6.x is very solid software, despite its age.  The newest version
> is (IMHO) better, but if 3.x (3.6.2 in particular) meets your needs, you
> can keep using it.  Solr 3.x can run with a very ancient version of Java
> -- version 5!  I believe that it still works even in Java 8.
>
>> I see that from 4.x to 5.x that Solr is now a standable server and no
>> longer just a WAR running on Tomcat.
>
> Yes.  There's a lot that could be said about that topic.  The highlights
> are here:
>
> https://wiki.apache.org/solr/WhyNoWar
>
> Thanks,
> Shawn
>

Re: [newbie] questions about 3.6.0 and 4.x or 5.x ?

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

On 10/21/2015 12:41 PM, Robert Hume wrote:
> I've inherited a project that uses a Solr 3.6.0 deployment.   (Several
> masters and several slaves – I think there are 6 Solr instances in total.)
>
> I've been tasked with investigating if upgrading our 3.6.0 deployment will
> improve performance – there’s a lot of data and things are getting slow,
> apparently.
>
> I've read Apache docs that from 3.6.x to 4.x there were improvements in
> scalability and performance.

Performance does get better in newer versions, but for most use cases,
there is NOT a night/day difference, just a minor speedup.  Upgrading
*might* help, but even if it does, chances are that it will not
completely solve the problem.

The most common reason for Solr performance problems is that there is
not enough memory.  That might mean the java heap is a little too small,
but more frequently, it means that there's not enough memory in the
server to cache the index contents effectively.

General information:

https://wiki.apache.org/solr/SolrPerformanceProblems

Solr 3.6.x is very solid software, despite its age.  The newest version
is (IMHO) better, but if 3.x (3.6.2 in particular) meets your needs, you
can keep using it.  Solr 3.x can run with a very ancient version of Java
-- version 5!  I believe that it still works even in Java 8.

> I see that from 4.x to 5.x that Solr is now a standable server and no
> longer just a WAR running on Tomcat.

Yes.  There's a lot that could be said about that topic.  The highlights
are here:

https://wiki.apache.org/solr/WhyNoWar

Thanks,
Shawn