Description of problem:
Running dbench on geo-rep setup with master config distributed-replicate and slave config distributed-replicat and parallely quota was being set and unset in loop. Dbench exited with "Bad file descriptor" 

100       686     4.19 MB/sec  execute 174 sec  latency 9912.135 ms
 100       686     4.16 MB/sec  execute 175 sec  latency 10343.618 ms
[656] write failed on handle 10011 (Bad file descriptor)
Child failed with status 1


Version-Release number of selected component (if applicable):3.3.0qa45


How reproducible:Frequently 


Steps to Reproduce:
1.Create geo-rep setup with master as distributed-replicate and slave as distribute-replicate 
2.Run dbench in the mount-point
3.enable and disable the quota in loop
  
Actual results:dbench exits with Bad file descriptor 


Expected results:dbench should complete properly.


Additional info:

volume info of the master 


Volume Name: master
Type: Distributed-Replicate
Volume ID: be06889e-66e2-4398-98be-a904652c7a42
Status: Started
Number of Bricks: 2 x 2 = 4
Transport-type: tcp
Bricks:
Brick1: 172.17.251.151:/exportdir/d1
Brick2: 172.17.251.151:/exportdir/d2
Brick3: 172.17.251.151:/exportdir/d3
Brick4: 172.17.251.151:/exportdir/d4
Options Reconfigured:
features.quota: on
geo-replication.indexing: on