Example |
 |
To simplify the example, this appendix uses a
generic eight-way (8-CPU) server whose CPUs are at the hypothetical
hardware paths of x01, x02, x03, x04, x05, x06, x07, and x08.
Create the Virtual Partitions
Suppose we create three partitions using the following
commands.
# vparcreate -p vpar1 -a cpu::2 -a cpu:::2
# vparcreate -p vpar2 -a cpu::2 -a cpu:::2
# vparcreate -p vpar3 -a cpu::1 |
And suppose that the vPars Monitor chooses the
following hardware paths for the bound CPUs:
Virtual
Partition | vpar1 | vpar2 | vpar3 |
Paths
of Bound CPU(s) | x01 x02 | x03 x04 | x05 |
This configuration leaves the following three
CPUs as unbound CPUs:
Paths
of Unbound CPUs | x06, x07, x08 |
Boot the Virtual Partitions
When we boot the partitions, they will boot with
the following bound CPUs; their respective kernels will have the following
unbound CPU entries.
Note that the entries for the unbound CPUs are
only entries for unbound CPUs that can potentially be added to the partition. At this point,
we have not assigned any unbound CPUs to any of the partitions:
Virtual
Partition | vpar1 | vpar2 | vpar3 |
Paths
of Bound CPU(s) | x01 x02 | x03 x04 | x05 |
Unbound
CPU Kernel Entries | x06 x07 x08 | x06 x07 x08 | x06 x07 x08 |
Paths
of Unbound CPUs | x06, x07, x08 |
Looking at vpar3, because
kernel entries for CPUs at x06, x07, and x08 exist, any of
the unbound CPUs (x06, x07, or x08) can be added
to vpar3. They could also be added to vpar1 or vpar2.
Create A Fourth Virtual Partition
Supposed we create and boot a fourth partition
using the following command:
# vparcreate -p vpar4 -a cpu::3 -a cpu:::3
The vPars Monitor will assign the remaining three
CPUs at hardware paths x06, x07, and x08:
Virtual
Partition | vpar4 |
Paths
of Bound CPU(s) | x06 x07 x08 |
Unbound
CPU Kernel Entries | (none) |
Remove a Virtual Partition
If we shutdown and remove vpar2 (using vparremove), its bound CPUs will become
unbound, and the current configuration will be the following:
Virtual
Partition | vpar1 | vpar3 | vpar4 |
Paths
of Bound CPU(s) | x01 x02 | x05 | x06 x07 x08 |
Unbound
CPU Kernel Entries | x06 x07 x08 | x06 x07 x08 | (none) |
Paths
of Unbound CPUs | unbound CPUs are now at x03 and x04 |
There are now two unbound CPUs, but
these CPUs are not the same ones that were available at the time the
partitions vpar1 or vpar3 were booted.
At this point, if we attempt to add an unbound
CPU to vpar3 using the following command:
# vparmodify -p vpar3 -a cpu::1
the command will fail and return the error message:
vparmodify Error: “-a cpu::1”:
One or more unbound CPUs were not available when virtual partition
vpar3 was booted. You must shutdown the partition to add them.
Although two unbound CPUs are available, their
hardware paths are x03 and x04. But the kernel entries for vpar3 are x06, x07, and x08. Therefore, the command will fail.
The Workaround: Reboot the Target Virtual Partition |
|
Because unbound CPU kernel entries are created
when the target partition is booted, you can reboot the target partition
so that kernel entries created correctly reflect the available unbound
CPUs.
In our example, if we want to add an unbound CPU
to vpar3, we can reboot vpar3:
vpar3# vparstatus
vpar3# shutdown -r |
When vpar3 boots again,
its kernel will create the correct entries for the unbound CPUs, which
are now at x03 and x04. The configuration becomes:
Virtual
Partition | vpar1 | vpar3 | vpar4 |
Paths
of Bound CPU(s) | x01 x02 | x05 | x06 x07 x08 |
Unbound
CPU Kernel Entries | x06 x07 x08 | x03 x04 | (none) |
Paths
of Unbound CPUs | unbound CPUs are now at x03 and x04 |
Because the unbound CPUs are at x03 and x04 and the kernel entries for vpar3 are x03 and x04, the command to add an unbound CPU to vpar3
# vparmodify -a vpar3 -a cpu::1
now will be successful.
Printable version
