We normally suggest to use HW RAID for data storage. This has the advantage that we can replace failed disk without service interruption (node stays online). When using sheepdog (or ceph), HW RAID additionally increase redundancy, and avoid recovery actions most times. Unfortunately, this is quite expensive compared to using a single disk (or 2 single disks without raid controller). With such single disk setup, is it possible to replace a damaged disk while the node keeps running (assuming HW supports hotplug)? NOTE: If we can't hotplug disks, any disk failure need a poweroff/reboot. That start auto-recovery and produces massive network traffic. If I run 2 sheeps on 2 separate disks (on same node), recovery use the second sheep to copy data (no network traffic involved)? Any experiences on those topic? - Dietmar -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.wpkg.org/pipermail/sheepdog-users/attachments/20120719/655398e7/attachment.html> |