SE35xx and SE33xx arrays - causes for slow LD rebuilds

Asset ID:	1-72-1020286.1
Update Date:	2009-10-24
Keywords:

Solution Type Problem Resolution Sure

Solution 1020286.1 : SE35xx and SE33xx arrays - causes for slow LD rebuilds

Related Items


Sun Storage 3510 FC Array
 Sun Storage 3310 Array
 Sun Storage 3511 SATA Array

Related Categories


GCS>Sun Microsystems>Storage - Disk>Modular Disk - 3xxx Arrays

PreviouslyPublishedAs
255568

Symptoms
After a disk failed in an SE35xx or ST33xx array (aka the 'minnow' array), the rebuild process seemed to take an extremely long time.

Resolution
One reason that the rebuilds can take a (relatively) long time on logicaldrives (LD) in these arrays can be related to the stripesize of the LD. The smaller the stripesize, the longer the rebuild will take.
For example, we have these two 1.09TB logical drives, each comprised of 4 disks in a RAID5 configuration:


LD    LD-ID        Size  Assigned  Type   Disks Spare  Failed Status
------------------------------------------------------------------------
ld0   0C62738A   1.09TB  Primary   RAID5  4     0      0      Good
                         Write-Policy Default          StripeSize  32KB

ld1   09FA2F26   1.09TB  Primary   RAID5  4     0      0      Good
                         Write-Policy Default          StripeSize 128KB

One LD has stripesize of 32KB, and the other has a stripesize of 128KB.

When we cause one drive to fail the LD with the larger (128KB) stripesize, it took this LD approximately 5 hours to rebuild to a hotspare.

When we cause one drive to fail the LD with the smaller (32KB) stripesize, it took this LD approximately 16 hours to rebuild to a hotspare.

Be aware that other factors can influence LD rebuild times, such as the amount of I/O to the array.

Product
Sun StorageTek 3310 SCSI Array
Sun StorageTek 3510 2U FC Array
Sun StorageTek 3510 FC Array
Sun StorageTek 3511 SATA Array

minnow, rebuild, GHS, hot spare, hotspare, slow rebuild, stripesize

Attachments

This solution has no attachment