Problem
I have successfully installed Virtual Disk Manager (VDM) and Release Supplement 1.1.0 (SCO OpenServer 5.0.2 ships with 1.1.0 installed). I have successfully configured my disk array but the system hangs when one of my disks fails. I expected Virtual Disk Manager to keep the system up until I was able to replace my disk.
Cause
The VDM software is a virtual disk driver which works between the application and operating system's filesystem code and the physical disks:
+------------------------------------------------+
| Application |
+------------------------------------------------+
| System Call Processing |
+------------------------------------------------+
| Filesystem, | Raw I/O |
| Buffer cache | Path |
+------------------------------------------------+
| Virtual Disk Driver |
+------------------------------------------------+
| Physical Disk Driver |
+------------------------------------------------+
| Physical Disk |
+------------------------------------------------+Having the system stay up when a drive fails, with the exception
of Raid 5 with a "hot spare," is hardware dependent. Notice the
Virtual Disk Driver sits on top of the physical layer, so if the
system is hanging due to problems at the hardware level, there
is nothing Virtual Disk Manager can do about it. There is hardware
available that can detect disk failure and allow the system to
stay up. Solution
Determine the needs of your system. If your system is to be a 24X7 server you will need hardware that will support these needs. You may want to look into a Hardware Raid Solution for such a requirement. If you do not have the necessary hardware and your system is hung due to disk failure, you must shutdown your system and replace the failed disk. You must then follow the proper steps to bring the disk array back online.
The proper steps will depend on which type of Raid you are using.
Raid 1-mirroring the root drive to another disk:
Follow the release notes for VDM Release Supplement 1.1.0, regarding replacing a mirrored root or parity disk.
Raid 5-without hot spare:Shut down the system and remove the failed disk. Replace it with a new disk at the same configuration as the previous one. If the previous was jumpered at id=4, for example, make sure the new one is jumpered at 4. Next, boot up to single-user mode and run mkdev hd to add the new disk. This will bring you into fdisk. You will fdisk the same way your previous disk was done. If you chose "Use entire disk for unix" previously, then do so again. You will then be asked whether or not you want to create a division table on the active partition. Choose "no" or the restore will fail. Upon completion, enter multiuser mode and parity will be restored to the new disk.
NOTE: Some disks may contain data VDM cannot work with. It may cause the parity restore to fail even though you followed these instructions. If SCSI, a low-level format prior to taking these steps will most likely prevent this problem.
Raid 10 or 53:Follow the same steps detailed in Raid 5 without hot spare.
See also
Chapter 8 in the System Administrators Guide and other related articles in this database.