[Qemu-devel] [PULL v3 00/41] Migration cleanups, refactorings, stats, and more
http://marc.info/?l=qemu-devel&m=135049188317185&w=2

- It introduces the stats discused previously
   * expected_downtime
   * downtime
   * dirty_pages_rate

This bug could be verified on qemu-kvm-rhev-1.5.3-31.el7.x86_64.

1. Migrate guest with the default configuration:

(qemu) info migrate
capabilities: xbzrle: off x-rdma-pin-all: off auto-converge: off zero-blocks: off 
Migration status: active
total time: 17943 milliseconds
expected downtime: 30 milliseconds  ==> Default downtime is correct.
setup: 149 milliseconds
transferred ram: 579095 kbytes
throughput: 268.57 mbps    ==> Match to the default migration speed 30MBps
remaining ram: 112640 kbytes
total ram: 4325856 kbytes
duplicate: 910810 pages
skipped: 0 pages
normal: 142494 pages
normal bytes: 569976 kbytes

(qemu) info migrate
capabilities: xbzrle: off x-rdma-pin-all: off auto-converge: off zero-blocks: off 
Migration status: completed
total time: 20572 milliseconds
downtime: 209 milliseconds
setup: 149 milliseconds
transferred ram: 660591 kbytes
throughput: 268.58 mbps
remaining ram: 0 kbytes
total ram: 4325856 kbytes
duplicate: 933137 pages
skipped: 0 pages
normal: 162779 pages
normal bytes: 651116 kbytes

Above values are updated during and after migration. 

2. Set "downtime" to 1s, "migration speed" to 100MBps:

(qemu) info migrate
capabilities: xbzrle: off x-rdma-pin-all: off auto-converge: off zero-blocks: off 
Migration status: active
total time: 4634 milliseconds
expected downtime: 1000 milliseconds  ==> Correct
setup: 59 milliseconds
transferred ram: 431412 kbytes
throughput: 830.57 mbps   ==> Correct
remaining ram: 279236 kbytes
total ram: 4325856 kbytes
duplicate: 905999 pages
skipped: 0 pages
normal: 105656 pages
normal bytes: 422624 kbytes

(qemu) info migrate
capabilities: xbzrle: off x-rdma-pin-all: off auto-converge: off zero-blocks: off 
Migration status: completed
total time: 6530 milliseconds
downtime: 258 milliseconds
setup: 59 milliseconds
transferred ram: 620975 kbytes
throughput: 830.88 mbps
remaining ram: 0 kbytes
total ram: 4325856 kbytes
duplicate: 935300 pages
skipped: 0 pages
normal: 152890 pages
normal bytes: 611560 kbytes

After migration, check the above values are reasonable. 


3. Turn on "xbzrle","auto-converge", "x-rdma-pin-all".

(qemu) info migrate
capabilities: xbzrle: on x-rdma-pin-all: on auto-converge: on zero-blocks: off 
Migration status: completed
total time: 19714 milliseconds
downtime: 205 milliseconds
setup: 54 milliseconds
transferred ram: 631236 kbytes
throughput: 268.57 mbps
remaining ram: 0 kbytes
total ram: 4325856 kbytes
duplicate: 936059 pages
skipped: 0 pages
normal: 155447 pages
normal bytes: 621788 kbytes
cache size: 67108864 bytes
xbzrle transferred: 5 kbytes
xbzrle pages: 90 pages
xbzrle cache miss: 5942
xbzrle overflow : 0


4. Test with qmp:

(1) default configuration:

{"execute":"query-migrate"}
{"return": {"expected-downtime": 30, "status": "active", "setup-time": 56, "total-time": 4867, "ram": {"total": 4429676544, "remaining": 4254449664, "mbps": 268.5668, "transferred": 164508924, "duplicate": 2701, "dirty-pages-rate": 0, "skipped": 0, "normal-bytes": 164163584, "normal": 40079}}

After migration finished:
{"return": {"status": "completed", "setup-time": 56, "downtime": 215, "total-time": 23002, "ram": {"total": 4429676544, "remaining": 0, "mbps": 268.5704, "transferred": 758725202, "duplicate": 905178, "dirty-pages-rate": 0, "skipped": 0, "normal-bytes": 749113344, "normal": 182889}}}

(2) with xbzrle:

{"return": {"xbzrle-cache": {"bytes": 660, "cache-size": 67108864, "pages": 21, "overflow": 0, "cache-miss": 2992}, "status": "completed", "setup-time": 57, "downtime": 211, "total-time": 23001, "ram": {"total": 4429676544, "remaining": 0, "mbps": 268.56968, "transferred": 757441588, "duplicate": 905213, "dirty-pages-rate": 0, "skipped": 0, "normal-bytes": 747831296, "normal": 182576}}}

So, the live migration statistics are included in the current qemu-kvm, this bug is fixed.

This request was resolved in Red Hat Enterprise Linux 7.0.

Contact your manager or support representative in case you have further questions about the request.