Partition Mobility is a function that has become critical to many customers environments to provide higher availability for planned outage such as firmware or server updates, workload balancing between servers and such. Because of this, we are frequently providing additional enhancements and this article will cover two recent enhancements.
Inactive partition migration with NPIV adapters
When using N_Port ID virtualization (NPIV), some of the information about the connection with the storage subsystem is retrieved when the connection is established. Because of this, in some situations the Virtual I/O server may not have enough information to perform all the tasks that could be performed as compared to a virtual SCSI connection. This was the case with inactive partition mobility. Inactive partition mobility is performing partition migration from one server to another server when the partition is powered off. Some users perform inactivate partition migration operations because they can be significantly faster than live partition mobility, especially on slow network connections. With inactive migrations, since the partition is powered off, the current contents of the partition’s memory do not need to be transmitted from the source server to the target server. In the past with NPIV storage, in order to perform partition migration, the very first migration operation for the partition needed to be an active migration. Once an active migration was performed for the partition, both inactive and active migrations were supported. A change has been made in all Power8 firmware levels and recently in FW770.70 and FW780.50 that allows an inactive partition migration operation to be the first migration operation thereby removing the need for the first migration to be an active migration.
Partition Migration support for affinity_group partition profile attributes
The PowerVM hypervisor, through the HMC command line interface, supports a couple of partition profile attributes that provide hints to the hypervisor about how the processor and memory resources should be assigned to a partition. These attributes are not used that frequently because the hypervisor normally is able to optimize the resource assignments based on the partition configuration. In the past, due to the specialized nature of these attributes, the attributes were ignored for partition migration requests. Through the Request For Enhancement (RFE) process, IBM received a request from a customer to honor the affinity_group attribute for migration requests. A change was made in HMC Release 8.6.0 SP 1 to migrate the affinity_group attribute from the source server to the target server. The usage of the RFE process is a great way for PowerVM customers to communicate their needs directly to the development teams.
Contacting the PowerVM Team
Have questions for the PowerVM team or want to learn more? Follow our discussion group on LinkedIn IBM PowerVM or IBM Community Discussions#PowerVM#powervmblog#powervmlpm