23.10.2014 Views

Advanced POWER Virtualization on IBM System p5 - Previous ...

Advanced POWER Virtualization on IBM System p5 - Previous ...

Advanced POWER Virtualization on IBM System p5 - Previous ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Virtual SCSI redundancy<br />

With the availability of MPIO <strong>on</strong> the client, each Virtual I/O Server can present a<br />

virtual SCSI device that is physically c<strong>on</strong>nected to the same physical disk. This<br />

achieves redundancy for the Virtual I/O Server itself and for any adapter, switch,<br />

or device that is used between the Virtual I/O Server and the disk.<br />

With the use of logical volume mirroring <strong>on</strong> the client, each Virtual I/O Server can<br />

present a virtual SCSI device that is physically c<strong>on</strong>nected to a different disk and<br />

then used in a normal AIX 5L mirrored volume group <strong>on</strong> the client. This achieves<br />

a potentially higher level of reliability by providing redundancy. Client volume<br />

group mirroring is also required when a Virtual I/O Server logical volume is used<br />

as a virtual SCSI device <strong>on</strong> the Client. In this case, the virtual SCSI devices are<br />

associated with different SCSI disks, each c<strong>on</strong>trolled by <strong>on</strong>e of the two Virtual I/O<br />

Server.<br />

Figure 4-1 <strong>on</strong> page 184 displays an advanced setup using both MPIO and LVM<br />

mirroring in the VIO client at the same time: two Virtual I/O Servers host disks for<br />

a single client. The client is using MPIO to access a SAN disk and LVM mirroring<br />

to access two SCSI disks. From the client perspective, the following situati<strong>on</strong>s<br />

could be handled without causing downtime for the client:<br />

►<br />

►<br />

►<br />

Either path to the SAN disk could fail, but the client would still be able to<br />

access the data <strong>on</strong> the SAN disk through the other path. No acti<strong>on</strong> has to be<br />

taken to reintegrate the failed path to the SAN disk after repair.<br />

The failure of a SCSI disk would cause stale partiti<strong>on</strong>s for the volume group<br />

with the assigned virtual disks, but the client still would be able to access the<br />

data <strong>on</strong> the other copy of these mirrored disk. After the repair of the failed<br />

SCSI disk or reboot of the Virtual I/O Server, all stale partiti<strong>on</strong>s would have to<br />

be synchr<strong>on</strong>ized using the vary<strong>on</strong>vg command.<br />

Either Virtual I/O Server could be rebooted, which would result in a temporary<br />

simultaneous failure of <strong>on</strong>e path to the SAN disk and stale partiti<strong>on</strong>s for the<br />

volume group <strong>on</strong> the SCSI disks, as described before.<br />

Chapter 4. Setting up virtual I/O: advanced 183

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!