SolrCloud boot with existing cores

0
votes

I am playing with solrcloud 4.6.0 and have a question. I have 3 standalone zookeeper v3.4.5 servers and 2 solr instances. Zookeeper was freshly installed and empty.

I configured the first solr node with the new discovery format, prepared initial existing cores (with data) and started the first solr with the bootstrap_conf set to true.

By observing the solr log I saw the cores config was uploaded to zookeeper and can also confirm that by looking at zookeeper registry. Also the cores are visible in the solr web console, searchable. All in all working.

Now I wanted the second solr node to kick in. The second solr was only having solr.xml (exactly the same as the first solr node). My understanding was that starting the second solr node will read the cores info from zookeeper and after a while all cores will replicate to it.

Didn't happen.

No errors in the logs, second solr web console says no cores are available.

What did I miss?

Best regards

1
solrapache-zookeepersolrcloud

1 Answers

0
votes

No errors in the log, huh? Bad eye I have.

It turns out that the initial configuration which was uploaded to zookeeper wasn't expending the variables in solr.xml, so the clusterstate.json had wrong ports for my Jetty solr installations - the default port 8983 was taken for replicas, but they were running on different ports.

I just modified the default value on all replicas in solr.xml file prior to booting the first solr instance with bootstrap_conf and now the cores are immediately visible. Also I deleted all zookeeper info before.

But still replicas have to be created on each node via web gui or by rest api.