We manage a large number of solr cores for a number of groups. To make this
manageable in production, we have a single 'multicore' configuration with all
cores deployed to a centralized NFS server, and each solr server loads this
shared configuration and writes its indexes to local disks. We have a number
of properties to turn on and off master/slave; and replication targets. This
lets us load certain indexes on certain machines entirely via JVM properties.
As the number of cores grows, we would like to be a bit more efficient and
only
load the cores that we will be using on each server. I was hoping to
accomplish this by enabling and disabling the cores via property subsititution
in solr.xml and then I can pass the properties as JVM parameters; but when I
tried this it appears that solr.xml doesn't respect the enable parameter that
works other places in solr. This is the configuration I tried:
<cores adminPath="/admin/cores">
<core name="group1" instanceDir="group1"
enable="${enable.core.group1:true}"/>
<core name="group2" instanceDir="group2"
enable="${enable.core.group2:true}"/>
<core name="group3" instanceDir="group3"
enable="${enable.core.group3:true}"/>
<core name="group4" instanceDir="group4"
enable="${enable.core.group4:true}"/>
<core name="group5" instanceDir="group5"
enable="${enable.core.group5:true}"/>
</cores>
If this isn't supported, another idea was to create multiple solr.xml files
(solr1.xml, solr2.xml, solr3.xml, solr4.xml, etc); where each xml file defines
the set of cores that we want to load; and then pass some JVM parameter which
tells the multicore which xml file to load at runtime. However, I can't find
any such parameter. Is the code only looking for "solr.xml" or is there some
secret parameter I can use to override?
Thanks
