Description of problem: ============================ On a 3 x 3 distribute-replicate volume with bricks on AWS-EC2 instances , added 3 more bricks to change the volume type to 4 x 3. Executed rebalance on the volume. After some time rebalance got successfully completed on all the nodes. Terminated the NODE3, NODE4, NODE8 and NODE12 Instances. Created 4 new instances and performed "replace-brick commit force" on each of the terminated nodes with the new nodes. Started "heal full" . While self-heal was in progress executed "peer detach force" on the terminated nodes. Also, "peer probed" 3 more new nodes. This started the rebalance process. Version-Release number of selected component (if applicable): =============================================================== glusterfs 3.4.0.57rhs built on Jan 13 2014 06:59:05 How reproducible: Steps to Reproduce: ======================= 1. Create a 3 x 3 distribute-replicate volume. Start the volume. 2. Create fuse mount. Create files/dirs from the mount point. 3. Add 3 more bricks to the volume. Start rebalance and wait for the rebalance to complete. 4. Poweroff few nodes (atleast 1 replica should be online) 5. Add new nodes to the pool. 6. Replace the bricks on the poweredoff node with the brick on new nodes. 7. Trigger "full" heal. 8. "peer detach force" the poweredoff nodes. 9. peer probe new nodes. Actual results: ========================= Peer probe starts rebalance process. Expected results: ===================== Peer probe should not start rebalance process. Additional info: ====================
Cloning this to 3.1. To be fixed in future.