Search <book_title>...

InfoScale™ 9.0 Virtualization Guide - Solaris

Last Published: 2025-04-14

Product(s): InfoScale & Storage Foundation (9.0)

Platform: Solaris

Section I. Overview of InfoScale solutions in Solaris virtualization environments
1. Overview of InfoScale virtualization solutions
Section II. Zones
1. InfoScale Enterprise Solutions support for Solaris Native Zones
Section III. Oracle VM Server for SPARC
Section IV. Reference
1. Appendix A. Where to find more information

Enabling DMP path failover in the guest domain

In Oracle VM Server configurations the Virtual Disk Client (VDC) driver timeout is set to zero by default which signifies infinity. This can cause the failed I/O not to return to the guest domain, if either the control or alternate I/O domain crashes unexpectedly. As a result the guest domain cannot get back the failed I/Os and cannot route them through the alternate domain. If this issue occurs or to avoid this issue, you must set the VDC driver timeout.

There are two ways to set the VDC driver timeout:

Modify globally all of the LUNs that are exported to the current guest domain. This requires a reboot to all the guest domains.	See “To change the VDC driver timeout globally”.
Manually export every LUN directly to the guest domain and set the timeout parameter to 30 seconds. No reboot is required.	See “To change the VDC driver timeout for each LUN”.

To change the VDC driver timeout globally

On each guest domain, edit the /etc/system file and add the following line to set the VDC driver timeout to 30 seconds:
```
set vdc:vdc_timeout=30
```
Reboot the guest domains.

To change the VDC driver timeout for each LUN

Create the primary domain using four internal disks and get all required SAN LUNs for the guest domains allocated to the primary domain.
Remove half of the system's I/O from the primary domain:
```
# ldm remove-io pci_X primary_domain_name
```
where pci_x is the name of the PCI bus for your system.
where primary_domain_name is the name of the primary domain.
For example:
```
# ldm remove-io pci_@400 primary
```
Create the alternate I/O domain on the other four internal disks and add the I/O that was removed from the primary domain:
```
# ldm add-io pci_X primary_domain_name
```
where pci_x is the name of the PCI bus for your system.
where primary_domain_name is the name of the primary domain.
For example:
```
# ldm add-io pci_@400 primary
```
On the primary domain, create the guest domains. In the sample the enclosure-based name of one of the LUNs is xyz and the guest domain is hsxd0015:
```
# ldm add-vdsdev /dev/vx/dmp/xyz vol0015-001-p1@primary-vds0
# ldm add-vdsdev /dev/vx/dmp/xyz vol0015-001-p2@alternate-vds0
# ldm add-vdisk timeout=30 vdsk0015-001-p1 \
vol0015-001-p1@primary-vds0 hsxd0015
# ldm add-vdisk timeout=30 vdsk0015-001-p2 \
vol0015-001-p2@alternate-vds0 hsxd0015
```
The same set of four commands for each SAN LUN that gets placed in a guest domain. Use three SAN LUNs for SAN boot in the guest domain and the rest for application data. Each LUN in the guest domain has one path backup through the primary domain and one backup through the alternate domain. That means each LUN only uses one LDC in each domain. Also, since you are using DMP you still only use 1 LDC in each domain, even if the LUN has more than two paths from the array.