IBM ServeRAID-8 Series, Лучшие практики и информация по обслуживанию

Страница: 31 / 58

v
The physical disks operate normally throughout the process.
v The array does not have an active background task (rebuild, copy back, or
initialize).
v
The array is not a RAID 10 or RAID 60.
This type of catastrophic failure is typically a symptom of other problems. There
must be two or more almost simultaneous failures (usually within milliseconds) for
the controller to fail in this way, and these problems are difficult to debug. Usually,
there is an obvious failure, but the secondary issue is often much more difficult to
detect. The best opportunity to get good data so that you can understand this
dynamic is to capture UART logs as soon as technically possible after the failure.
You can collect UART logs by typing the following CLI command:
arcconf getlogs n uart > filename.txt
Valid RAID 5EE state transitions after a disk fails
RAID 5EE has some unique characteristics. This RAID level uses a distributed
hot-spare technique that populates free space throughout the array. If a disk fails,
the controller compacts the missing data into the free space of the surviving drives.
This activity creates a new substate of the logical drive, called
compacting . These
substates identify the stages of the ServeRAID controllers as they recover from a
defunct drive. The RAID 5EE valid state transitions for single-disk failures are listed
in the following table.
Table 2. RAID 5EE valid state transitions for single-disk failures
Logical drive state RAID 5EE state Description
Okay Expanded
v
Normal and Optimal
Critical Compacting
v A single drive is defunct.
v The array is actively compacting to free
space.
v
The array is degraded in performance.
v
The array is not redundant at this time.
Critical Compacted
v A single drive is defunct.
v The array has compacted to the free space.
v The array is not degraded in performance.
v The array is redundant.
Critical Expanding
v
The defunct drive was replaced.
v The array is expanding to its original
configuration.
v
The array is degraded in performance.
v
The array is redundant.
At the time that a RAID 5EE state becomes Critical, the array should start
compacting shortly thereafter. Drives do not rebuild in a RAID 5EE until the array
has finished compacting, even if the drive is replaced before the compacting is
completed.
If the array does not start compacting automatically, the controller might report the
array as a Critical array with a substate as Expanded. Unfortunately, this
combination of states is not valid. Typically, an array in this condition remains
Chapter 1. ServeRAID-8 series best practices and maintenance information
25