On 2010/01/08 7:10, Piavlo wrote: > Hi, > > I've noticed that after stopping all sheepdogs, I need to start them in > the same order as then the cluster was setup for > the first time (or maybe I just need to start first the node one which > mkfs was run - not sure) > otherwise sheepdogs do not communicate: > > fire-srv4 ~ # shepherd info -t dog > Idx Node id (FNV-1a) - Host:Port > -------------------------------------------------- > fire-srv4 ~ # > fire-srv4 ~ # shepherd info -t sheep > Id Size Used Use% > > Total 0G 0G -2147483648%, total virtual VDI Size 0G > fire-srv4 ~ # > > While they do communicate at corosync level: > > fire-srv4 ~ # corosync-cpgtool > Group Name PID Node ID > sheepdog > 3031 53364868 (192.1.1.3) > 28593 86919300 (192.1.1.5) > 15526 70142084 (1921.1.4) > fire-srv4 ~ # > > after stopping sheepdogs and restarting in the "RIGHT" order storage > works again. Sorry, rebooting sheepdog cluster is not supported yet. It is one of top priority TODOs and will be done next week. Thanks, Kazutaka Morita |