At Sat, 20 Aug 2011 16:22:17 +0200, Valerio Pachera wrote: > > From getting started: > 'It is enough to run this command on one of machines. "--copies" > specifies the number of default data redundancy. In this case, the > replicated data will be stored on three machines.' > > Correct me if I'm wrong: using --copies=3, I have always at least 3 > copies for any chunck of a vdi disk. > So, if my scluster uses 5 or more nodes, when one of them dies, the > chunks stored on will be moved to the other nodes. Yes. > > The chunks are distrubted on all nodes. So I do not have 3 nodes with > data and 2 without. > > If I add a node (the 6th) simply by "sheep /speed_data", the chinks > will automatically be distrbuted also on the new node. Yes. > > If I shutdown a node after the other, giving it the time necessesary > for data sync, all the data will be move to the last node (off course > with enough space). > There will be no redudance and the last node will not have any 'extra' > chunk copy left. As you may know, you should run "collie cluster shutdown" before stopping a Sheepdog cluster to avoid extra data copies.. > > In this schenario, one node "exports" one directory, so removing a > node equals to remve a "cluster member". Anyway, a single node can > export more directories so ti would be more correct to speak about > "removing directory form the cluster"? I'm not sure this answers your questions, but currently you need to run multiple sheep daemons to handle multiple directories. Thanks, Kazutaka |