On Fri, Jan 08, 2016 at 15:56:02 +0100, Peter Krempa wrote:
On Fri, Jan 08, 2016 at 10:49:37 +0100, Jiri Denemark wrote:
memory_dirty_rate corresponds to dirty-pages-rate in QEMU and memory_iteration is what QEMU reports in dirty-sync-count.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Looks more like a feature rather than just "cleanup"
--- include/libvirt/libvirt-domain.h | 19 +++++++++++++++++++ src/qemu/qemu_domain.c | 8 ++++++++ src/qemu/qemu_migration.c | 12 ++++++++++++ src/qemu/qemu_monitor.h | 2 ++ src/qemu/qemu_monitor_json.c | 4 ++++ tools/virsh-domain.c | 16 ++++++++++++++++ 6 files changed, 61 insertions(+)
diff --git a/include/libvirt/libvirt-domain.h b/include/libvirt/libvirt-domain.h index a1ea6a5..d26faa5 100644 --- a/include/libvirt/libvirt-domain.h +++ b/include/libvirt/libvirt-domain.h @@ -2724,6 +2724,25 @@ int virDomainAbortJob(virDomainPtr dom); */ # define VIR_DOMAIN_JOB_MEMORY_BPS "memory_bps"
+/** VIR_DOMAIN_JOB_MEMORY_DIRTY_RATE: + * + * virDomainGetJobStats field: number of memory pages dirtied by the guest + * per second, as VIR_TYPED_PARAM_ULLONG. This statistics makes sense only + * when live migration is running.
Do we document somewhere how to convert page sizes into memory sizes (or rather how big the pages are/whether hugepages count as 1 or more etc...?)
No we don't :-) And QEMU doesn't document that either. Honestly I doubt anyone would even try to convert this into memory sizes. It's enough to monitor the trend to see whether a migration is converging or not. If the value is decreasing, there is a chance migration will eventually finish, otherwise, it's obvious that the guest is dirtying more memory than QEMU can transfer to the destination host.
+ */ +# define VIR_DOMAIN_JOB_MEMORY_DIRTY_RATE "memory_dirty_rate" + +/** + * VIR_DOMAIN_JOB_MEMORY_ITERATION: + * + * virDomainGetJobStats field: current iteration over domain's memory + * during live migration, as VIR_TYPED_PARAM_ULLONG. This is set to zero + * when memory starts to be transferred and the value is increased by one + * every time a new iteration is started to transfer memory pages dirtied + * since the last iteration. + */ +# define VIR_DOMAIN_JOB_MEMORY_ITERATION "memory_iteration"
ACK
I pushed this series, thanks for the review. Jirka