I think ZK stuff may actually be easier to handle, no? Add new ones to the existing ZK cluster and then remove the old ones. Won't this work smoothly?
Otis -- Monitoring * Alerting * Anomaly Detection * Centralized Log Management Solr & Elasticsearch Support * http://sematext.com/ On Thu, Oct 30, 2014 at 1:16 PM, Jakov Sosic <jso...@gmail.com> wrote: > On 10/30/2014 04:47 AM, Otis Gospodnetic wrote: > >> Hi/Bok Jakov, >> >> 2) sounds good to me. It means no down-time. 1) means stoppage. If >> stoppage is not OK, but falling behind with indexing new content is OK, >> you >> could: >> * add a new cluster >> * start reading from old index and indexing into the new index >> * stop old cluster when done >> * index new content to new cluster (or maybe you can be doing this all >> along if indexing old + new at the same time is OK for you) >> -- >> > > Thank you for suggestions Otis. > > Everything is acceptable currently, but in the future as the data grows, > we will certainly enter those edge cases where neither stopping indexing > nor stopping queries will be acceptable. > > What makes things a little bit more problematic is that ZooKeepers are > migrating also to new machines. > > >
