About Data Protection In A Single-Controller Storage System - HP P2000 Reference Manual

RAID-6 reconstruction behaves as follows:
• If one disk fails and a compatible spare is available, the system begins to use that spare to reconstruct
the vdisk. If a second disk fails during reconstruction, reconstruction continues until it is complete,
regardless of whether a second spare is available. If the spare fails during reconstruction,
reconstruction stops.
• If two disks fail and only one compatible spare is available, the system waits five minutes for a second
spare to become available. After five minutes, the system begins to use that spare to reconstruct one
disk in the vdisk (referred to as "fail 2, fix 1" mode). If the spare fails during reconstruction,
reconstruction stops.
• If two disks fail and two compatible spares are available, the system uses both spares to reconstruct the
vdisk. If one of the spares fails during reconstruction, reconstruction proceeds in "fail 2, fix 1" mode. If
the second spare fails during reconstruction, reconstruction stops.
• If a disk fails during online initialization, the initialization fails. In order to generate the two sets of
parity that RAID 6 requires, the controller fails a second disk in the vdisk, which changes the vdisk
status to critical, and then assigns that disk as a spare for the vdisk. If a second spare is not available
within five minutes, reconstruction proceeds in "fail 2, fix 1" mode.
When a disk fails, its Fault/UID LED illuminates amber. When a spare is used as a reconstruction target, its
Online/Activity LED illuminates green.
NOTE: Reconstruction can take hours or days to complete, depending on the vdisk RAID level and size,
disk speed, utility priority, and other processes running on the storage system. You can stop reconstruction
only by deleting the vdisk.

About data protection in a single-controller storage system

A P2000 G3 MSA System can be purchased or operated with a single controller. Because single-controller
mode is not a redundant configuration, this section presents some considerations concerning data
protection.
A volume's default caching mode is write back, as opposed to write through. In write-back mode, data is
held in controller cache until it is written to disk. In write-through mode, data is written directly to disk.
If the controller fails while in write-back mode, unwritten cache data likely exists. The same is true if the
controller enclosure or the target volume's enclosure is powered off without a proper shut down. Data
remains in the controller's cache and associated volumes will be missing that data. This can result in data
loss or in some cases volume loss; for example, if using snapshot functionality a snap pool might become
inaccessible and the master volume could go offline.
If the controller can be brought back online long enough to perform a proper shut down, the controller
should be able to write its cache to disk without causing data loss.
If the controller cannot be brought back online long enough to write its cache data to disk, you can move
its CompactFlash cache card to a replacement controller. This enables the cache data to be available
when the new controller comes online. The CompactFlash card is externally accessible from the back of the
controller.
To avoid the possibility of data loss in case the controller fails you can change a volume's caching mode to
write through. While this will cause significant performance degradation, this configuration guards against
data loss. While write-back mode is much faster, this mode is not guaranteed against data loss in the case
of a controller failure. If data protection is more important, use write-through caching; if performance is
more important, use write-back caching.
For details about caching modes see
caching mode, see
34 Getting started
About volume cache options
Changing a volume's cache settings
on page 24. To change a volume's
on page 55.