Heidi Eckhart wrote:
Kaitlin Rupert wrote:
Is the type CIM_MIGRATE_LIVE or VIR_MIGRATE_LIVE ? The CIM_MIGRATE_LIVE corresponds to the value defined in the mof, which is 2. However, the value of VIR_MIGRATE_LIVE is defined by
Heidi Eckhart wrote: libvirt.
So when someone specifies a migration type of 2 (CIM_MIGRATE_LIVE) in the MigrationSettingData, we'll need to use VIR_MIGRATE_LIVE when we call virDomainMigrate(). Thanks for this explanation.
For static migration, libvirt uses a value of 0. There's two type values here.. the value set for the MigrationType attribute of MigrationSettingData and the value libvirt uses to differentiate between static and live migration. ... It is a little more redundant this way, but it might help with readability.
Thank you for this very good explanation. This makes now much more sense to me :). But I think recoding as you suggested is not necessary, as it would bring too much redundancy (as you already said) into the code. Now that I understood it, I'm absolutely fine with it. Its not very consistent that libvirt uses VIR_MIGRATE_LIVE and 0 (instead of VIR_MIGRATE_STATIC or something else) as values, but that's something we seem to live with.
Yes, this confused me too as well. =) When I checked libvirt.h, I was surprised to see only one migration flag listed. I wonder if this will change in the future? -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com