Dima,

Thanks for the reply!  However, this does not quite answer my question, as far 
as I can tell.  I am very familiar with network proxies, and have both Nginx 
proxy (externally facing) and Apache (internal load balancing) on my network.  
I am comfortable with how to distribute search queries across nodes.

My fundamental question — and sorry if this was not clear — is how to I keep 
the indices (collections) in-sync across nodes?  Put another way, if I update 
shard1 on one node, how do I get the other node(s) automatically updated?  The 
goal is to be able to do indexing on a particular node, and have any updates 
propagate across the other nodes, so that the indices (collections) are 
identical (hopefully within a few seconds) across all of the nodes.

Of course, one way is to have a shared filesystem to share the index 
(collection) data files across all of the nodes … but then the shared 
filesystem becomes a single point of failure.

It appears that Solr knows how to replicate the indices (collections) across 
nodes, so that there is no single point of failure.  This is what I am trying 
to figure out.

Thanks,

Dave.

> On Oct 20, 2023, at 11:52 AM, Dmitri Maziuk <dmitri.maz...@gmail.com> wrote:
> 
> On 10/19/23 18:48, David Filip wrote:
> 
>> My goal is to replicate content from one to the other, so that I can take 
>> one down (e.g., solr1) and still search current collections (e.g., on solr2).
> 
> You need a proxy host, it can be anything from apache to F5, configured to 
> pass requests to Solr nodes, based on some criteria.
> 
> In the active-passive, blue-green, or whatever you call it, configuration, 
> you and don't need zookeeper or anything shared on the backend (there is an 
> argument for having the backend nodes fully independent).
> 
> If you RTFM: see Query Fault Tolerance" in 
> https://solr.apache.org/guide/solr/latest/deployment-guide/solrcloud-distributed-requests.html
>  -- even if you use SolrCloud you still need a proxy for what you want done. 
> (Unless your client application knows how to talk to zookeper and can use it 
> as the proxy.)
> 
> As an aside, it's interesting that Apache httpd does not have a mod_zookeper 
> among its proxy modules.
> 
> Dima
> 

Reply via email to