Zk holds more than just the Solr config, it holds a copy of the clusterstate, which includes all the sharding, hash ranges, etc as well. You will need to re-create that data on your new ZK instance, i.e. re-create the collection to populate that.
You do realize that running a single ZK instance is relatively dangerous though (for production at least)? If that ZK dies or is shutdown, you have no failover, and your cloud will become read-only (since Solr won't be able to find the leader for any shard) On 20 May 2016 at 17:50, Scott Chu <scott....@udngroup.com> wrote: > > Even worse when I startup Solrclouds nodes and point back the original > co-op zk nodes. It shows can't load conf for shard 2. It seems pointing to > standalone zk node previously already hurt some config data so that the > Solrcloud nodes can no long talk to original co-op zk nodes perfectly. Is > it possible to repair the config data? > > Scott Chu,scott....@udngroup.com > 2016/5/21 (週六) > ----- Original Message ----- > From: scott.chu > To: solr-user > CC: > Date: 2016/5/21 (週六) 00:44 > Subject: Switching zk node cause load conf error > > > > I intially start up 3 zk nodes and upload config 'cugna'. Then I start 2 > Solrcloud nodes, create collection with 2 shards, add a lot of docs ok. > Later I find 3 zk nodes occupy too many CPU and memory, so I start a > standalone zk node and use Solr's zkcli to upload same config 'cugna'. > This time I start same 2 Solrcloud nodes but point to this standalone zk > node. They seem to start ok. > However, when I open localhost:8983, it shows can't load config for shard1 > and shard2. > Does this mean I can't switch other zk nodes once after I create index? > The same config on co-op and standalone seem to have difference? > > Scott Chu,scott....@udngroup.com > 2016/5/21 (週六) > > > ----- > 未在此訊息中找到病毒。 > 已透過 AVG 檢查 - www.avg.com > 版本: 2015.0.6201 / 病毒庫: 4568/12265 - 發佈日期: 05/20/16 >