Hi, Lon (2011/10/26 9:54), Lon Hohberger wrote: > On 09/20/2011 03:40 AM, Kazunori INOUE wrote: >> Hi, >> >> I created the backend "pm-fence" which has met the following >> configurations / requirements. >> - Both hosts and VMs, cluster (Pacemaker) have been configured. >> * This backend has STONITH (fencing) function of the VM cluster. >> - The VMs are managed as resources (RA is VirtualDomain) by the host cluster. >> * In this configuration, in order to stop a VM, >> it is necessary to stop the resource through the host cluster. >> When the VM was stopped without through the host cluster >> (e.g., run the 'virsh destory'), the host cluster considers that >> VM resource failed. >> > > This patch is good, but the server/Makefile.in needs to use pkg-config > --cflags / --libs glib-2.0 rather than static definitions. I have > included an updated patch. > OK. But please correct the target of 'pkg-config -libs' to ncurses.
# diff 0001-backend-plugin-for-Pacemaker.patch mod 64c64 < +PACEMAKER_LIBS=-lcib -lpe_status `pkg-config --libs glib-2.0` --- > +PACEMAKER_LIBS=-lcib -lpe_status `pkg-config --libs ncurses` # When glib-2.0 is the target, it fail in load. # pkg-config --libs glib-2.0 -lglib-2.0 # # fence_virtd -d99 -F : Loading plugin from /DEV/fence-virt/plugins/pm-fence.so Could not dlopen /DEV/fence-virt/plugins/pm-fence.so: /usr/lib64/libpe_status.so.2: undefined symbol: stdscr > Otherwise, I do like this patch quite a lot. > > The only question I had is why there needs to be an external STONITH > plugin; I was of the impression that fence_virt and fence_xvm work out > of the box with Pacemaker 1.1.x. That doesn't affect this particular We do not know how to do it... > patch, as far as I can tell (it looks like just a glue wrapper between > pacemaker and fence_virt itself). > That's right. The purpose of this STONITH plugin is to adapt the request from a pacemaker to usage of fence_virt/fence_xvm. Best Regards, Kazunori INOUE
