Shawn - this worked great! I reloaded the collection, the nodes recovered and it seems to work great now.
Thank you, Aki On Thu, Feb 4, 2016 at 7:10 PM, Shawn Heisey <apa...@elyograg.org> wrote: > On 2/4/2016 2:12 PM, Aki Balogh wrote: > > I found the state.json file and it indeed shows that the range for shard1 > > is null. > > > > In order to fix, do I need to upload a corrected state.json file with > > corrected hash ranges? How can I do that? (zkcli.sh?) > > The easiest way to figure out the correct hash range, if you do not know > it and cannot figure it out by looking at the other hash ranges, is to > create a new collection with the same number of shards as the broken > collection, then look at the clusterstate for that collection to see > what the hash ranges are, to determine which range is missing. Then > once you have fixed the existing collection, delete the new collection. > > You can upload the changed state.json file with zkcli.sh, I think the > command is "putfile", and you will need the full path within zookeeper. > Alternately, you can get a GUI zookeeper client for IDEs like Eclipse > and IntelliJ IDEA for a more interactive experience. Reloading the > collection after replacing the state.json file is probably required. > > Thanks, > Shawn > >
