Creating a duplicate mapping throws the whole cluster

[clintongormley]

Hiya

I'm starting 3 nodes, then creating a mapping, then creating a duplicate (with ignoreDuplicate=false).

It is sufficient to throw the whole cluster out. It doesn't seem to recover.

Run this script a few times, and watch the server logs:

#!/bin/bash
curl -XGET 'http://127.0.0.1:9200/_cluster/nodes' 
curl -XDELETE 'http://127.0.0.2:9202/es_test_1/' 
curl -XPUT 'http://127.0.0.2:9202/es_test_1,es_test_2/test/_mapping?ignoreDuplicates=false'  -d '
{
   "properties" : {
      "num" : {
         "type" : "integer"
      },
      "text" : {
         "type" : "string"
      }
   }
}
'
curl -XPUT 'http://127.0.0.2:9202/es_test_1,es_test_2/test/_mapping?ignoreDuplicates=false'  -d '
{
   "properties" : {
      "num" : {
         "type" : "integer"
      },
      "text" : {
         "type" : "string"
      }
   }
}
'
curl -XPUT 'http://127.0.0.2:9202/es_test_1,es_test_2/test_2/_mapping?ignoreDuplicates=false'  -d '
{
   "properties" : {
      "num" : {
         "type" : "integer"
      },
      "text" : {
         "type" : "string"
      }
   }
}
'
curl -XDELETE 'http://127.0.0.2:9202/es_test_1/' 
curl -XDELETE 'http://127.0.0.2:9202/es_test_2/' 
curl -XPUT 'http://127.0.0.2:9202/es_test_1/'  -d '
{}
'
curl -XPUT 'http://127.0.0.2:9202/es_test_2/'  -d '
{}
'
curl -XPUT 'http://127.0.0.2:9202/_all/type_1/_mapping?ignoreDuplicates=false'  -d '
{
   "properties" : {
      "num" : {
         "store" : "yes",
         "type" : "integer"
      },
      "text" : {
         "store" : "yes",
         "type" : "string"
      }
   }
}
'
curl -XPUT 'http://127.0.0.2:9202/_all/type_2/_mapping?ignoreDuplicates=false'  -d '
{
   "properties" : {
      "num" : {
         "store" : "yes",
         "type" : "integer"
      },
      "text" : {
         "store" : "yes",
         "type" : "string"
      }
   }
}
'
curl -XPOST 'http://127.0.0.2:9202/_flush?refresh=true' 

[kimchy]

Fixed one issue, where the proper exception was not raised (actually serialized) since you have ignoreDuplicates to false. Trying to recreate the problem you posted.

[kimchy]

Hopefully, its fixed now. Since the allocation of shards is async, the create/delete/create/delete cycle caused the delete api to hand in the ack phase. Can you give it a go? (note, I added a cluster health API where you can wait for the cluster to get to a certain state)

[clintongormley]

This looks like it has been fixed