On Tue, 2012-10-30 at 02:06 -0700, Tarek Ziadé wrote:
On Tuesday, October 30, 2012 3:56:23 AM UTC+1, Vineeth Mohan wrote: 
A better option would be to maintain aliases - 
Elasticsearch Platform — Find real-time answers at scale | Elastic 
    Thanks
              Vineeth
 
That's what we do now - but I have an existing instance with some 
indexes I wanted to copy over.
 
Ah, you can import dangling indices into a cluster.  The problem is when 
they have the same name as existing indices.
I opened this issue about exactly that problem:
  
  
    
  
  
    
    
      
        opened 12:41PM - 10 Oct 12 UTC 
      
        
          closed 05:01PM - 10 Nov 12 UTC 
        
      
     
    
    
   
 
  
    If I copy the directory that contains an index to a new directory name, then res… tart the cluster, I see the message:
```
[INFO ][gateway.local.state.meta ] [Sytsevich, Aleksei] dangled index directory name is [new_name], state name is [old_name], renaming to directory name
```
This, to me, implies that the state name is being changed to the directory name.  In other words, I should have a new index called `new_name`.
Actually, it just ignores the index because we already have `old_name`.
If I edit the state file to change the state name, then I get a new index with the edited name.
So first: the log message is inaccurate. It should be "renaming to state name".
But second, I think the better behaviour would be to update the state name to the directory name.  For one, this would make it easy to physically rename an index. 
   
   
  
    
    
  
  
 
clint
 
--