Enhanced support (9.7 and 10.1) Live Migration Support Status & Roadmap

• • Backward migration from major release: 10 -> 9
    • [Steven] Confirmed with mdeng@redhat.com,  x86 includes this part of the testing

RHEL 9.6.0 <-> RHEL 10.0 (Test Scenario: ping-pong migration and post-copy migration)

• [dzheng] libvirt level also provides migration test from 9->10->9

• Precopy
• TLS encryption
• XBZRLE (lower prio than TLS, most probably not used by Kubevirt)
• [dzheng] confirmed with libvirt pre-integration with CNV guy, CNV does not use this option

• Dirty rate monitoring (calc-dirty-rate)
• Auto-converge
• Xiaohui: Auto-converge provides a method for dynamically throttling guest cpus during autoconverge migrations in order to force migration of very busy guests to complete. (see AutoconvergeLiveMigration for details)
• Refer to case: VIRT-48421 This case is also supported on ARM

• multifd
• Core (basic support)
• TLS (multifd on TLS channels)
• XBZRLE
• + auto-converge (multifd running concurrently with auto-converge feature enabled)
• Zerocopy (multifd can send pages with reduced memcpy, using MSG_ZEROCOPY socket interface)

• Postcopy
• core
• TLS encryption
• XBZRLE (lower prio)
• recovery (xiaohli@redhat.com: tested by using migrate-pause qmp cmd to pause postcopy on ARM, not with real network)
• preemption

• Virtiofs live migration (what is the x86 status), sync with VFIO team (this is not tested by the migration team)
• [Steven]  I have discussed with Dan.This part is already supported on x86. Dan will confirm the part related to ARM and update it later.Dan Zheng
• [dzheng] x86_64 has automated tests for it, we can enable them on ARM

• Out of scope:

Multifd + postcopy (RHEL-58189, RHEL-59697): According to peter, this is still under work upstream. Not yet sure this will land in 9.7/10.1