In contrast to DR operations that take place immediately, delayed reconfiguration operations take effect in the following circumstances:
After the next reboot of the OS
After a stop and start of a logical domain if no OS is running
In general, delayed reconfiguration operations are restricted to the control domain. For all other domains, you must stop the domain to modify the configuration unless the resource can be dynamically reconfigured.
Delayed reconfiguration operations are restricted to the control domain. You can run a limited number of commands while a delayed reconfiguration on the root domain is in progress to support operations that cannot be completed dynamically. These subcommands are add-io, set-io, remove-io, create-vf, and destroy-vf. You can also run the ldm start-reconf command on the root domain. For all other domains, you must stop the domain to modify the configuration unless the resource can be dynamically reconfigured.
While a delayed reconfiguration is in progress, other reconfiguration requests for that domain are deferred until it is rebooted or stopped and started.
The ldm cancel-reconf command cancels delayed reconfiguration operations on the domain. For more information about how to use the delayed reconfiguration feature, see the ldm(1M) man page.
You can use delayed reconfiguration to decrease resources on the control domain. To remove a large number of CPUs from the control domain, see Removing a Large Number of CPUs From a Domain Might Fail in Oracle VM Server for SPARC 3.2 Release Notes . To remove large amounts of memory from the control domain, see Decrease the Control Domain's Memory.
Do not attempt to perform more than one CPU configuration operation on the primary domain while it is in a delayed reconfiguration. If you attempt more CPU configuration requests, they will be rejected.
Workaround: Perform one of the following actions:
Cancel the delayed reconfiguration, start another one, and request the configuration changes that were lost from the previous delayed reconfiguration.
Reboot the control domain with the incorrect CPU count and then make the allocation corrections after the domain reboots.