Sun Fire T2000 Server
 

GCS>Sun Microsystems>Servers>CMT Servers
 

***Checked for relevance on 17-May-2011***

Applies to:

Goal

Solution

• Integrated Stripe or IS volumes (RAID0)
• Integrated Mirror or IM volumes (RAID1)

Before using the Hardware RAID features of the Sun Fire T2000 server, ensure that the following patches have been applied:

• 119850-22 (or later) -  mpt and raidctl patch, this patch maybe obsoleted, for the latest mpt patch refer to My Oracle Support(MOS) - Patches and Updates
  
• 122165-02  Provides LSI1064 updated FCode to the PCI-X SAS disk controller card (required only for earlier T2000 systems using this card)

Note: To confirm the installed PCI-X FCode revision see the patch README. The ALOM showhost command cannot be used as this command queries the onboard (currently inactive) LSI chip firmware and will not show the correct version.

NOTE: Mirroring the boot disks is typically done prior installing the OS since the RAID controller requires usage of 64MB of each disk in the RAID. This must be completed using the procedure documented in the "Sun Fire T2000 Server Product Notes" which is the top link from:

Sun SPARC Enterprise T2000 Server Documentation

This document describes hardware RAID functionality, practical usage, and specifications on the Niagara-based Sun Fire T2000 server. The Sun Fire T2000 server currently utilizes an LSI SAS1064 chipset and a 4 PHY-wide SAS HBA with up to four (4) 72G 2.5" SAS disk drives as well as support for RAID0 (Striping) and RAID1 (Mirroring).

SAS (Serial Attached SCSI) provides:

• A serial, full-duplex point-to-point topology.
• Up to 3Gb/s per port.
• Smaller form factor hard disk drives.
• Advanced command queuing.
• Full interface compatibility with SATA components.

The two Solaris[TM] Operating System components of Sun Fire T2000 RAID functionality are:

  -mpt:    Solaris kernel module for support of LSI Logic MPI-enabled HBAs.
There is current support for the LSI1030 and LSI1064 HBAs.

  -raidctl(1M):  Solaris utility to interface with the built-in RAID
functionality of MPI HBAs. 

RAID levels available:

RAID1: Mirroring

• A 2-disk primary->secondary configuration.
• Upon creation, secondary is populated with primary's data (sync).
• All write I/Os are committed to both disks before returning to provide for data redundancy.
• Read I/Os may be serviced by either disk.
• When a single disk fails, data is still available, but the volume is in "DEGRADED" mode. The failed disk may be replaced and resync'ed.
• Once sync is complete, mirroring may be disabled/broken to provide two (2) separate drives with the same exact data.

RAID0: Striping

• A 2 or more disk configuration.
• No data is shared. Data is across all concatenated disks in exactly one (1) location.
• Write I/Os are committed in a round-robin fashion across all member disks. Performance is increased in some workload scenarios.
• Read I/Os are satisfied at their one location, performing as a single disk.
• Upon a single disk failure, data is lost.
• If a RAID0 volume is disabled/broken, data is lost.

Operations Summary:

Create a mirror:

      raidctl -c [-r 1] primary secondary

Create a stripe:

      raidctl -c -r 0 disk1 disk2 [disk3] . . .

Delete a RAID volume:

      raidctl -d volume

Display RAID volumes:

      raidctl

Update HBA controller firmware:

      raidctl -F  [path_of_image_file] [controller#] 

RAID Operation Details

Creating a mirror

Format:

      raidctl -c [-r 1] primary secondary

Example:

      # raidctl -c c1t0d0 c1t1d0

With the above command issued, a RAID volume c1t0d0 will be created. Data on c1t0d0 will be synced to c1t1d0, which will overwrite data on c1t1d0. No file systems may be mounted on the secondary disk. The secondary disk must be as large as the primary disk for successful creation. The RAID volume will end up having the total capacity of just one disk.

Creating a stripe

Format:

      raidctl -c -r 0 disk1 disk2 [disk3] . . .

Example:

      # raidctl -c -r 0 c1t1d0 c1t2d0 c1t3d0

When the above command is issued, RAID volume c1t1d0 will be created. Data on all the drives will be destroyed, and no file systems may be mounted on any individual disk. The RAID volume will have have the capacity of all the disks combined (The smallest common size is used).

Deleting a RAID volume

Format:

      raidctl -d volume

Example:

      # raidctl -d c0t0d0

When the above command is issued, the RAID volume will be deleted. Data may be lost depending on the RAID configuration. Please see the RAID descriptions above.

Checking RAID status

Example:

      # raidctl

Example output:

    RAID      RAID      RAID      Disk
Volume    Status    Disk      Status
----------------------------------------
c0t0d0    RESYNCING c0t0d0    OK
c0t1d0    OK

The volume (RAID) status can be one of four possible:

• OK: Volume is optimal.
• DEGRADED: Volume contains a failed disk (mirror) or is running with reduced functionality.
• RESYNCING: Volume is synchronizing (mirror), disks are sync'ing.
• FAILED: Volume is non-functional/offline.

The disk status can be one of three possible:

• OK: Disk is operating normally.
• MISSING: Disk is removed or otherwise offline.
• FAILED: Disk is in a failed state and needs service.

After RAID volume creation

• Member disk drives are no longer visible, only the RAID volume is presented.
• The RAID volume may be addressed just as any other LUN in Solaris.
• The status of the RAID volume and its member disks may be retrieved using the raidctl(1M) command.
• LSI1064 HBA supports up to 2 RAID volumes at any one time.
• After RAID volume creation a valid Solaris label must be applied so that the new volume can be accessed correctly by the OS. Please see for detailed steps.

This solution has no attachment