When you say you move to different machines, did you copy the zoo_data from your old setup, or did you just start up zookeeper and your shards one by one? Also did you use collection API to create the collection or just start up your cores and let them attach to ZK. I believe the ZK rules for assigning shards has changed somewhere around 4.2. We had a setup with 4.0 and it simply assigned them in order, shard 1, shard 2, shard 3, etc then when all shards were filled, it started with replicas.
In 4.3 (we skipped the intermediates) the ordering wasn't obvious, I had to do a bit of trial an error to determine the right order to start things in order to get shard assignments correct, but that isn't really the recommended way of doing it. If you want specific assignments (cores to shards) then I think the core API/collection API are the recommended way to go. Create a collection using the Collection API (http://wiki.apache.org/solr/SolrCloud) and then copy the data to the right servers once it has assigned the shards (it should make sure that replicas don't exist on the same machine, and things like that). I believe the general direction (of the next major Solr release) is to start a system with a blank solr.xml and create cores/collections that way rather than have a file and then have to connect to ZK and merge the data with what's there. We have a slightly odd requirement in that we need to determine the DataDir for each core, and I haven't yet worked out the right sequence of commands (Collection API doesn't support DataDir but CoreAPI does). It should be possible though, just haven't found the time to get to it! On 25 June 2013 18:40, Erick Erickson <erickerick...@gmail.com> wrote: > Try sending requests to your shards with &distrib=false. See if the > results agree with the SolrCloud graph or whether the docs you get > back are inconsistent with the shard labels in the admin page. The > &distrib=false bit keeps the query from going to other shards and > will tell you if the current state is consistent or not..... > > Best > Erick > > On Tue, Jun 25, 2013 at 1:02 AM, Shalin Shekhar Mangar > <shalinman...@gmail.com> wrote: > > Firstly, using 1 zookeeper machine is not at all ideal. See > > http://wiki.apache.org/hadoop/ZooKeeper/FAQ#A7 > > > > I've never personally seen such an issue. Can you give screen shots of > > the cloud graph on each node? Use an image hosting service because the > > mailing list won't allow attachments. > > > > On Tue, Jun 18, 2013 at 2:07 PM, Ophir Michaeli <micha...@wesee.com> > wrote: > >> Hi, > >> > >> I built a 2 shards and 2 replicas system that works ok on a local > machine, 1 > >> zookeeper on shard 1. > >> It appears ok on the solar monitor page, cloud tab > >> (http://localhost:8983/solr/#/~cloud). > >> When I move to using different machines, each shard/replica on a > different > >> machine I get a wrong cloud-graph on the Solr monitoring page. > >> The machine that has Shard 2 appears on the graph on shard 1, and the > >> replicas are also mixed, shard 2 appears as 1 and shard 1 appears as 2. > >> > >> Any ideas why this happens? > >> > >> Thanks, > >> Ophir > > > > > > > > -- > > Regards, > > Shalin Shekhar Mangar. >