lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shawn Heisey <>
Subject Re: High Availability with two nodes
Date Tue, 26 Feb 2019 17:17:16 GMT
On 2/26/2019 2:39 AM, Andreas Mock wrote:
> currently we are looking at Apache Solr as a solution
> for searching. One important component is high availability.
> I digged around finding out that HA is built in via
> SolrCloud which means I have to install ZooKeeper in
> a production environment which needs at least three
> nodes.

ZK requires at least three nodes.  This is how it is designed, and the 
only way that is is possible to be absolutely certain that operation is 

The third node (running ZK only, not Solr) does not need much power. 
You could get a very cheap laptop for that role and it would probably 
work great.

> So, now to my problem. I haven't found any documents
> showing a way to get a two node cluster simply for
> HA (active/passive).
> Is there a recommended way or are there any solutions
> out there showing a scenario with Solr Single Server
> combined e.g. DRBD and Pacemaker?

SolrCloud cannot achieve this with two servers, due to ZK requirements.

You could set up two servers using master/slave replication ... but if 
your master dies, indexing becomes a challenge.  Switching which machine 
is master requires manual reconfiguration of all servers ... something 
that wouldn't be the case with SolrCloud ... because it does not have 
masters and slaves.  The leader for each index is transitory, assigned 
by a ZK election.


View raw message