First, looks at the admin/cloud tab and make sure the urls are right and reachable for each node. Are you using 4.0 or 4.1? 4.1 should make a better guess at the host name to use - you can override that and be explicit if it's guessing wrong. If you have nodes on different machines, you don't want it to be localhost.
Next, look at the logs. They should give a clue why the replicas can't recover from the leader. - Mark On Feb 27, 2013, at 8:25 AM, Christopher Gross <cogr...@gmail.com> wrote: > I've been trying out Solr 4 -- I was able to get it working with 3 > instances of Tomcat on the same box (different ports), and 5 Zookeeper > nodes on that box as well. I've started to get my production layout going, > but I can't seem to get the Solr to replicate among the nodes. > > I can see that the Solr instance in each Tomcat gets hooked into the > Zookeeper (I can see items get added into the > (collections/leader_elect/shard1/election tree part of the "Cloud" admin > page), but only the first node that I start up will get the data, it won't > replicate to the other places. > > I copied my working configurations (from the everything on the same server > setup), changing the hostnames and directories to point to the right spot. > > I find it disconcerting that Zookeeper has "localhost" for each of the > nodes getting hooked in, I'd feel better if it was saying the actual > hostname for the servers that are connecting, but I'm not sure if that's > the issue or not. > > I can give log or config file snippets, if that will help. They're > identical among the systems, outside of any hostnames that are specific to > the server. I do have iptables running, but it still didn't work when I > had that shut off, so I don't believe it to be a port issue (has the ZK & > Tomcat ports open for in/out anyway). > > Has anyone else run into anything like this? Any suggestions of where to > look? I'm thinking that it's just a small config setting somewhere at this > point, but it all looks OK to me. > > -- Chris
