I am experimenting with an approach. My current plan is to run Cinder LVM on the compute node and attach block device of the local cinder LVM backend directly to the VM. The cinder LVM backend would be a VG span across JBOD disks. Anybody ever tried this? Are there any special directive required so that Cinder-scheduler knows that the local Cinder LVM backend should be used? On Sun, May 10, 2015 at 10:51 AM, Jagat Singh <jagatsingh at gmail.com> wrote: > We are running Hadoop on Openstack. But have disks configured as RAID5. We > get terrible disk throughput. > > How do we do openstack disk configuration for JBOD in the way Hadoop likes. > > We are running Cisco UCS device , each server has 32 cores and 10 disks > but on Openstack Hadoop layer in datanode each disk appears as one at this > moment. > > Is there any documentation or best practices around configuring Openstack > for Hadoop. I know there is Savana project but that does not shares how to > do things at physical level. > > Thanks > > _______________________________________________ > Mailing list: > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > Post to : openstack at lists.openstack.org > Unsubscribe : > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.openstack.org/pipermail/openstack/attachments/20150510/2ddea02a/attachment.html>