Hi, I'm using the SBD fencing method in my Career Final Project. I'm student but work like linux sysadmin since years ago. :)
In Ubuntu Krusty, i don't find the package for bsd and the bsd software was removed from cluster-glue package. Compile and install bsd was pretty hard but it was done. The sbd agent shows the follow information: --- root@nodo1:/usr/lib/stonith/plugins/external # ./sbd getinfo-devdescr sbd uses a shared storage device as a medium to communicate fencing requests. This allows clusters without network power switches; the downside is that access to the shared storage device becomes a Single Point of Failure. It requires sbd to be configured on all nodes. Please read http://linux-ha.org/wiki/SBD_Fencing! --- I don't understand why the SBD method is a SPOF. If it use a SAN, the block device can be in mirror RAID in the Storage. With multipath fibre gets redundancy in the communications. Where would be the SPOF? it would be the sbd daemon? it would be reasonable to run the bsd daemon like resource in Pacemaker? I think controld agent resource might be right to run the bsd daemon. I think the I/O fencing is the best option because it doesn't depend of network to be operative. It is SBD truely independent from network status for work? I have my dudes because in the compilation of the binary, sbd needs corosync and pacemaker libs. Finally, IMHO, let me some suggestions to improve SBD. It would be nice that by default, "sbd list", "sbd dump" and others commands , would work without having to specify the device parameters. Of course, the variable SBD_DEVICE in /etc/sysconfig/sbd file, would be defined. --- root@nodo1:/usr/lib/stonith/plugins/external # grep SBD_DEVICE /etc/sysconfig/sbd # SBD_DEVICE specifies the devices to use for exchanging sbd messages SBD_DEVICE="/dev/disk/by-id/ata-VBOX_HARDDISK_VB37c86fb9-b67e3db4-part1;/dev/disk/by-id/ata-VBOX_HARDDISK_VB24ac6909-20d73ce9-part1;/dev/disk/by-id/ata-VBOX_HARDDISK_VB051291f4-f95e847d-part1" root@nodo1:/usr/lib/stonith/plugins/external # sbd list You must specify 1 to 3 devices via the -d option. sbd failed; please check the logs. root@nodo1:/usr/lib/stonith/plugins/external # sbd -d /dev/disk/by-id/ata-VBOX_HARDDISK_VB37c86fb9-b67e3db4-part1 -d /dev/disk/by-id/ata-VBOX_HARDDISK_VB24ac6909-20d73ce9-part1 -d /dev/disk/by-id/ata-VBOX_HARDDISK_VB051291f4-f95e847d-part1 list 0 nodo2 clear 1 nodo1 clear 0 nodo2 clear 1 nodo1 clear 0 nodo2 clear 1 nodo1 clear --- In debian and ubunut distros, the /etc/sysconfig directory doesn exist. The directory equivalent would be /etc/defaults. The initscript sbd is not compatible in debian distros. I modified the provieded script to work in ubuntu and debian. I don't know if that would be task for the debian HA maintainers. I think that the initscript should be load the softdog module in the start function if the parameter is defined. One last question. Which is the official repository and website of the SBD proyect? - http://linux-ha.org/wiki/SBD_Fencing - https://github.com/l-mb/sbd/ - http://hg.linux-ha.org/sbd/ Thank you so much! A great and cool software, Corosync, Pacemaker and SBD. Saludos.
_______________________________________________ Pacemaker mailing list: Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
