I think I found the solution using self-healing. Before that I need to make an image of the running node as the recovery image to save the self-healing time. Do you guys have other ideas about it? Yueyu, Lin Sent from my iPad On May 2, 2011, at 1:47 PM, Yueyu Lin <yueyu.lin at me.com> wrote: > > Hi, all > Maybe the title is not that clear. Let me explain my scenario in details here. > Assuming I have a storage cluster in distributed replication configuration. The replica factor is 2 that means every brick has another one to be backup for each other. > Whenever one of the brick is crashed, it will be detected by our system immediately. Our system can recover it using another CLEAN system. CLEAN means all the configurations including MAC and IP are the same. But the new system has no data stored in it at all. > After the new instance is restarted, since it has the same IP, it will be added to the cluster automatically. But the problem is that it has no data, will the GlusterFS find the data inconsistency and replicate from another brick to the new recovered brick? Or should I run fschk to trigger the data fixing? > Thanks a lot in advance. > > Yueyu, Lin > Sent from my iPad > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
