For 3ware SATA RAID controllers, initialize means to put the redundant data on the drives of redundant units into a known state so that data can be recovered in the event of a disk drive failure. This is sometimes referred to as
background initialization or
resynching, and does not erase user data.
Some RAID levels must be initialized for best performance. (For specifics, see
Initialization of Different RAID Types.) When these units are created in the BIOS (through 3BM), you can choose to do a
foreground initialization, which will take place before the operating system has loaded, or a
background initialization, which allows you to put the unit in service immediately, but will slow down the unit performance until it completes. You can elect to cancel foreground initialization, put the units into service, and have initialization run in the background, instead.
|
|
Mac User Note: Foreground initializations are only done in 3BM, which is not applicable for Mac users.
|
Initialization makes parity information valid. Foreground initialization does this by simply writing zeroes to all the drives so that they all have the same values, overwriting any existing data in the process. In contrast, background initialization uses an algorithm to resynch the parity information on the drives and does not rewrite existing data.
A foreground initialization can be run from the BIOS using 3BM. It clears all existing data from the drives. Foreground initialization can take up to several hours, depending upon the size of the unit. After foreground initialization completes, you can start the operating system and units will perform at peak efficiency.
|
|
Note: Drives that support the Write Same feature (SCT) can write to multiple drive sectors at once, improving initialization time. To take advantage of this feature, all the drives in the unit must support Write Same.
|
If you are creating a unit through 3BM and if immediate access to the unit is important, select background initialization. Background initialization will then start automatically within about 10 minutes. Background initialization (resynching) does not write zeroes to the drives or harm your data. You can partition, format, and use the unit safely while it is initializing. The unit is fully fault-tolerant while the initialization takes place. That is, if the unit degrades before the initialization is complete, the data will remain intact.
When initializing is done after booting to the operating system, the process of initializing takes longer than it does if initialization is done by writing zeroes to the unit in the BIOS. Consequently, it will be a longer period of time until the performance of the unit is fully optimal.
Although you can use the unit while it is being initialized in the background, initialization does slow I/O performance until completed. You can adjust how much initialization will slow performance by setting the rate at which it occurs. (See
Setting Background Task Rate.) You can also postpone initialization until a scheduled time. (See
Scheduling Background Tasks).
|
|
Note: Units that are not initialized through 3BM are automatically initialized using background initialization when they are verified for the first time. (Verification requires that the units have been previously initialized.) This will not affect the data on the drives, and the units will perform normally, although performance will be slowed until the initialization and verification are completed.
|
RAID 0 units do not need to be initialized and cannot be initialized. RAID 0 units are immediately available for use with full performance when created.
Regardless of the size, all 9000-series RAID 5, RAID 6, and RAID 50 units are fully fault tolerant upon creation. These configurations use a specialized scheme for writing to the unit, which does not have to be valid to provide fault tolerance.
Performance of the following types of units will improve after the unit has been initialized: RAID 6, RAID 5 units with 5 or more disks, and RAID 50 units with 2 subunits of 5 or 6 disks. For these configurations, initialization begins automatically after you create them. If you create them in 3BM using the default foreground initialization method, zeroes are written to all unit members before booting to the operating system. If you create them through 3DM2, you will boot to the operating system and the RAID 5 parity is then calculated and written to disk, keeping any data in the unit intact.
RAID 5 units with 3 or 4 disks do not need to be initialized to have full performance upon creation. It is okay that 3 or 4 disk RAID 5 units are not initialized. These RAID types are fully redundant, regardless of whether or not they are initialized. Similarly, RAID 50 units with a grouping of 3 or 4 disks in a subunit do not need to be initialized. However, RAID 50 with a grouping of 5 or more disks in a subunit do need to be initialized for full performance.
RAID 1 and RAID 10 units do not need to be initialized when they are created. They are fault tolerant and are immediately available for use with full performance when created.
Initialization of a RAID 1 unit results in data from one disk (the disk on the lower port number) being copied to the other disk. In RAID 10 units, data from one half of the unit is copied to the other half.
After the initialization, subsequent verifies to a RAID 1 or RAID 10 unit check for data consistency by comparing the data from one drive (or set of drives) to the other drive (or set of drives).
The 3ware controller detects and handles power failures, using a mechanism that ensures that redundant units have consistent data and parity. When a redundant unit is unexpectedly shutdown, there is a possibility some data and parity may be inconsistent. If a unit or sub-unit of a redundant unit is detected to have been shutdown uncleanly, the unit or sub-unit will change its mode to either ‘Initializing’ or ‘Verifying.’
