Re: [Qemu-devel] [PULL 00/34] migration thread and queue

[Alexandre DERUMIER]

Hi,
I'm currently testing new migration code with last qemu.git,

it's working pretty fine (around 30ms downtime with standard workload).


But I have add some problem, with high memory workload vm. (playing an HD video 
for example).

Target vm is pause after migration,
# info status
VM status: paused (internal-error)

(downtime is around 700ms)


I can reproduce it 100%

Regards,

Alexandre Derumier


----- Mail original -----

De: "Juan Quintela" <address@hidden>
À: address@hidden
Cc: address@hidden
Envoyé: Vendredi 21 Décembre 2012 20:41:03
Objet: [Qemu-devel] [PULL 00/34] migration thread and queue

Hi

Changes for last version:
- inlined paolo fixes, attached here (they were very small, and only change 
error cases locking)
- I tested with autotest and see no problems

Please pull.

Thanks, Juan.


diff --git a/arch_init.c b/arch_init.c
index 86f8544..ea75441 100644
--- a/arch_init.c
+++ b/arch_init.c
@@ -641,13 +641,13 @@ static int ram_save_iterate(QEMUFile *f, void *opaque)
}
i++;
}
+ qemu_mutex_unlock_ramlist();

if (ret < 0) {
bytes_transferred += total_sent;
return ret;
}

- qemu_mutex_unlock_ramlist();
qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
total_sent += 8;
bytes_transferred += total_sent;
@@ -657,9 +657,8 @@ static int ram_save_iterate(QEMUFile *f, void *opaque) 

static int ram_save_complete(QEMUFile *f, void *opaque)
{
- migration_bitmap_sync();
-
qemu_mutex_lock_ramlist();
+ migration_bitmap_sync();

/* try transferring iterative blocks of memory */

diff --git a/migration.c b/migration.c
index c69e864..d6ec3e8 100644
--- a/migration.c
+++ b/migration.c
@@ -650,7 +650,7 @@ static int64_t buffered_set_rate_limit(void *opaque, 
int64_t new_rate)
new_rate = SIZE_MAX;
}

- s->xfer_limit = new_rate / 10;
+ s->xfer_limit = new_rate / XFER_LIMIT_RATIO;

out:
return s->xfer_limit;


[20121220]
Changes for yesterday:
- Paolo Acked the series
- Rebase on top of today git (only conflicts were due to header re-shuffle)

Please pull.

[20121219]

This is my queue for migration-thread and patches associated. This
integrates review comments & code for Paolo. This is the subset from
both approachs that we agreed with. rest of patches need more review
and are not here.

Migrating and idle guest with upstwream:

(qemu) info migrate
capabilities: xbzrle: off
Migration status: completed
total time: 34251 milliseconds
downtime: 492 milliseconds
transferred ram: 762458 kbytes
remaining ram: 0 kbytes
total ram: 14688768 kbytes
duplicate: 3492606 pages
normal: 189762 pages
normal bytes: 759048 kbytes

with this series of patches.

(qemu) info migrate
capabilities: xbzrle: off
Migration status: completed
total time: 30712 milliseconds
downtime: 29 milliseconds
transferred ram: 738857 kbytes
remaining ram: 0 kbytes
total ram: 14688768 kbytes
duplicate: 3503423 pages
normal: 176671 pages
normal bytes: 706684 kbytes

Notice the big difference in downtime. And that is also seen inside
the guest a program that just do an idle loop seeing how "long" it
takes to wait for 10ms.

with upstream:

address@hidden ~]# ./timer
delay of 452 ms
delay of 114 ms
delay of 136 ms
delay of 135 ms
delay of 136 ms
delay of 131 ms
delay of 134 ms

with this series of patches, wait never takes 100ms, nothing is printed.

Please pull.

Thanks, Juan.

The following changes since commit 27dd7730582be85c7d4f680f5f71146629809c86:

Merge remote-tracking branch 'bonzini/header-dirs' into staging (2012-12-19 
17:15:39 -0600)

are available in the git repository at:


git://repo.or.cz/qemu/quintela.git thread.next

for you to fetch changes up to 381c08083929f50f4780ea272ea36f7e5899b3b6:

migration: merge QEMUFileBuffered into MigrationState (2012-12-21 20:01:24 
+0100)

----------------------------------------------------------------
Juan Quintela (25):
migration: include qemu-file.h
migration-fd: remove duplicate include
buffered_file: Move from using a timer to use a thread
migration: make qemu_fopen_ops_buffered() return void
migration: move migration thread init code to migrate_fd_put_ready
migration: make writes blocking
migration: remove unfreeze logic
migration: just lock migrate_fd_put_ready
buffered_file: Unfold the trick to restart generating migration data
buffered_file: don't flush on put buffer
buffered_file: unfold buffered_append in buffered_put_buffer
savevm: New save live migration method: pending
migration: move buffered_file.c code into migration.c
migration: add XFER_LIMIT_RATIO
migration: move migration_fd_put_ready()
migration: Inline qemu_fopen_ops_buffered into migrate_fd_connect
migration: move migration notifier
ram: rename last_block to last_seen_block
ram: Add last_sent_block
memory: introduce memory_region_test_and_clear_dirty
ram: Use memory_region_test_and_clear_dirty
ram: optimize migration bitmap walking
ram: account the amount of transferred ram better
ram: refactor ram_save_block() return value
migration: merge QEMUFileBuffered into MigrationState

Paolo Bonzini (7):
migration: fix migration_bitmap leak
buffered_file: do not send more than s->bytes_xfer bytes per tick
migration: remove double call to migrate_fd_close
exec: change ramlist from MRU order to a 1-item cache
exec: change RAM list to a TAILQ
exec: sort the memory from biggest to smallest
migration: fix qemu_get_fd for BufferedFile

Umesh Deshpande (2):
add a version number to ram_list
protect the ramlist with a separate mutex

Makefile.objs | 3 +-
arch_init.c | 244 +++++++++++++-------------
block-migration.c | 49 ++----
buffered_file.c | 268 -----------------------------
buffered_file.h | 22 ---
dump.c | 8 +-
exec.c | 128 +++++++++-----
include/exec/cpu-all.h | 15 +-
include/exec/memory.h | 16 ++
include/migration/migration.h | 13 +-
include/migration/qemu-file.h | 5 -
include/migration/vmstate.h | 1 +
include/sysemu/sysemu.h | 1 +
memory.c | 16 ++
memory_mapping.c | 4 +-
migration-exec.c | 3 +-
migration-fd.c | 4 +-
migration-tcp.c | 3 +-
migration-unix.c | 3 +-
migration.c | 390 +++++++++++++++++++++++++++++++-----------
savevm.c | 24 ++-
target-i386/arch_dump.c | 2 +-
22 files changed, 599 insertions(+), 623 deletions(-)
delete mode 100644 buffered_file.c
delete mode 100644 buffered_file.h