Hi, ring has 2 copies (5 zones, 1 zone = 1 node), we want to remove a machine that is faulty if we remove all devices of this one node from ring and disconnect this node from LAN (not waiting ending replication), then replication completed success?
Sorry for my english, thanks.
Thanks! One node working not stable, we want replace it with another mashine. but there is no have extra slots in the rack. We think make replace (remove old and add new) on one rebalance.
If you update the ring (remove the faulty device and rebalance) and then deploy the new version of the ring to the remaining 4 servers, replication will ensure that the data that was on the faulty device will be fully replicated from the one remaining copy.
Note that because you are running with only 2 replicas and you have a faulty machine, there is some subset of your data that currently only has a single copy. This is very dangerous from a durability perspective, and you should update your ring as quickly as possible to ensure you get back to your 2 replicas. (As a side note, I'd also recommend that you use 3 replicas instead of 2 for data that you cannot afford to lose.)
Asked: 2013-04-26 05:22:37 -0500
Seen: 33 times
Last updated: Apr 26 '13
