* [PULL 2/5] migration/throttle: Add throttle-trig-thres migration parameter
2020-03-13 10:30 [PULL 0/5] migration queue Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 1/5] configure: Improve zstd test Dr. David Alan Gilbert (git)
@ 2020-03-13 10:30 ` Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 3/5] COLO: Optimize memory back-up process Dr. David Alan Gilbert (git)
` (3 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: Dr. David Alan Gilbert (git) @ 2020-03-13 10:30 UTC (permalink / raw)
To: qemu-devel, zhang.zhanghailiang, zhukeqian1, quintela
From: Keqian Zhu <zhukeqian1@huawei.com>
Currently, if the bytes_dirty_period is more than the 50% of
bytes_xfer_period, we start or increase throttling.
If we make this percentage higher, then we can tolerate higher
dirty rate during migration, which means less impact on guest.
The side effect of higher percentage is longer migration time.
We can make this parameter configurable to switch between mig-
ration time first or guest performance first.
The default value is 50 and valid range is 1 to 100.
Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com>
Message-Id: <20200224023142.39360-1-zhukeqian1@huawei.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
---
migration/migration.c | 24 ++++++++++++++++++++
migration/ram.c | 52 +++++++++++++++++++++++++------------------
monitor/hmp-cmds.c | 7 ++++++
qapi/migration.json | 16 ++++++++++++-
4 files changed, 76 insertions(+), 23 deletions(-)
diff --git a/migration/migration.c b/migration/migration.c
index 0b2045ccbd..59da5d0625 100644
--- a/migration/migration.c
+++ b/migration/migration.c
@@ -78,6 +78,7 @@
/*0: means nocompress, 1: best speed, ... 9: best compress ratio */
#define DEFAULT_MIGRATE_COMPRESS_LEVEL 1
/* Define default autoconverge cpu throttle migration parameters */
+#define DEFAULT_MIGRATE_THROTTLE_TRIGGER_THRESHOLD 50
#define DEFAULT_MIGRATE_CPU_THROTTLE_INITIAL 20
#define DEFAULT_MIGRATE_CPU_THROTTLE_INCREMENT 10
#define DEFAULT_MIGRATE_MAX_CPU_THROTTLE 99
@@ -778,6 +779,8 @@ MigrationParameters *qmp_query_migrate_parameters(Error **errp)
params->compress_wait_thread = s->parameters.compress_wait_thread;
params->has_decompress_threads = true;
params->decompress_threads = s->parameters.decompress_threads;
+ params->has_throttle_trigger_threshold = true;
+ params->throttle_trigger_threshold = s->parameters.throttle_trigger_threshold;
params->has_cpu_throttle_initial = true;
params->cpu_throttle_initial = s->parameters.cpu_throttle_initial;
params->has_cpu_throttle_increment = true;
@@ -1169,6 +1172,15 @@ static bool migrate_params_check(MigrationParameters *params, Error **errp)
return false;
}
+ if (params->has_throttle_trigger_threshold &&
+ (params->throttle_trigger_threshold < 1 ||
+ params->throttle_trigger_threshold > 100)) {
+ error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
+ "throttle_trigger_threshold",
+ "an integer in the range of 1 to 100");
+ return false;
+ }
+
if (params->has_cpu_throttle_initial &&
(params->cpu_throttle_initial < 1 ||
params->cpu_throttle_initial > 99)) {
@@ -1298,6 +1310,10 @@ static void migrate_params_test_apply(MigrateSetParameters *params,
dest->decompress_threads = params->decompress_threads;
}
+ if (params->has_throttle_trigger_threshold) {
+ dest->throttle_trigger_threshold = params->throttle_trigger_threshold;
+ }
+
if (params->has_cpu_throttle_initial) {
dest->cpu_throttle_initial = params->cpu_throttle_initial;
}
@@ -1382,6 +1398,10 @@ static void migrate_params_apply(MigrateSetParameters *params, Error **errp)
s->parameters.decompress_threads = params->decompress_threads;
}
+ if (params->has_throttle_trigger_threshold) {
+ s->parameters.throttle_trigger_threshold = params->throttle_trigger_threshold;
+ }
+
if (params->has_cpu_throttle_initial) {
s->parameters.cpu_throttle_initial = params->cpu_throttle_initial;
}
@@ -3558,6 +3578,9 @@ static Property migration_properties[] = {
DEFINE_PROP_UINT8("x-decompress-threads", MigrationState,
parameters.decompress_threads,
DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT),
+ DEFINE_PROP_UINT8("x-throttle-trigger-threshold", MigrationState,
+ parameters.throttle_trigger_threshold,
+ DEFAULT_MIGRATE_THROTTLE_TRIGGER_THRESHOLD),
DEFINE_PROP_UINT8("x-cpu-throttle-initial", MigrationState,
parameters.cpu_throttle_initial,
DEFAULT_MIGRATE_CPU_THROTTLE_INITIAL),
@@ -3667,6 +3690,7 @@ static void migration_instance_init(Object *obj)
params->has_compress_level = true;
params->has_compress_threads = true;
params->has_decompress_threads = true;
+ params->has_throttle_trigger_threshold = true;
params->has_cpu_throttle_initial = true;
params->has_cpu_throttle_increment = true;
params->has_max_bandwidth = true;
diff --git a/migration/ram.c b/migration/ram.c
index 0ef68798d2..02cfd76d19 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -896,11 +896,38 @@ static void migration_update_rates(RAMState *rs, int64_t end_time)
}
}
+static void migration_trigger_throttle(RAMState *rs)
+{
+ MigrationState *s = migrate_get_current();
+ uint64_t threshold = s->parameters.throttle_trigger_threshold;
+
+ uint64_t bytes_xfer_period = ram_counters.transferred - rs->bytes_xfer_prev;
+ uint64_t bytes_dirty_period = rs->num_dirty_pages_period * TARGET_PAGE_SIZE;
+ uint64_t bytes_dirty_threshold = bytes_xfer_period * threshold / 100;
+
+ /* During block migration the auto-converge logic incorrectly detects
+ * that ram migration makes no progress. Avoid this by disabling the
+ * throttling logic during the bulk phase of block migration. */
+ if (migrate_auto_converge() && !blk_mig_bulk_active()) {
+ /* The following detection logic can be refined later. For now:
+ Check to see if the ratio between dirtied bytes and the approx.
+ amount of bytes that just got transferred since the last time
+ we were in this routine reaches the threshold. If that happens
+ twice, start or increase throttling. */
+
+ if ((bytes_dirty_period > bytes_dirty_threshold) &&
+ (++rs->dirty_rate_high_cnt >= 2)) {
+ trace_migration_throttle();
+ rs->dirty_rate_high_cnt = 0;
+ mig_throttle_guest_down();
+ }
+ }
+}
+
static void migration_bitmap_sync(RAMState *rs)
{
RAMBlock *block;
int64_t end_time;
- uint64_t bytes_xfer_now;
ram_counters.dirty_sync_count++;
@@ -927,26 +954,7 @@ static void migration_bitmap_sync(RAMState *rs)
/* more than 1 second = 1000 millisecons */
if (end_time > rs->time_last_bitmap_sync + 1000) {
- bytes_xfer_now = ram_counters.transferred;
-
- /* During block migration the auto-converge logic incorrectly detects
- * that ram migration makes no progress. Avoid this by disabling the
- * throttling logic during the bulk phase of block migration. */
- if (migrate_auto_converge() && !blk_mig_bulk_active()) {
- /* The following detection logic can be refined later. For now:
- Check to see if the dirtied bytes is 50% more than the approx.
- amount of bytes that just got transferred since the last time we
- were in this routine. If that happens twice, start or increase
- throttling */
-
- if ((rs->num_dirty_pages_period * TARGET_PAGE_SIZE >
- (bytes_xfer_now - rs->bytes_xfer_prev) / 2) &&
- (++rs->dirty_rate_high_cnt >= 2)) {
- trace_migration_throttle();
- rs->dirty_rate_high_cnt = 0;
- mig_throttle_guest_down();
- }
- }
+ migration_trigger_throttle(rs);
migration_update_rates(rs, end_time);
@@ -955,7 +963,7 @@ static void migration_bitmap_sync(RAMState *rs)
/* reset period counters */
rs->time_last_bitmap_sync = end_time;
rs->num_dirty_pages_period = 0;
- rs->bytes_xfer_prev = bytes_xfer_now;
+ rs->bytes_xfer_prev = ram_counters.transferred;
}
if (migrate_use_events()) {
qapi_event_send_migration_pass(ram_counters.dirty_sync_count);
diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
index 1c69d51b97..58724031ea 100644
--- a/monitor/hmp-cmds.c
+++ b/monitor/hmp-cmds.c
@@ -407,6 +407,10 @@ void hmp_info_migrate_parameters(Monitor *mon, const QDict *qdict)
monitor_printf(mon, "%s: %u\n",
MigrationParameter_str(MIGRATION_PARAMETER_DECOMPRESS_THREADS),
params->decompress_threads);
+ assert(params->has_throttle_trigger_threshold);
+ monitor_printf(mon, "%s: %u\n",
+ MigrationParameter_str(MIGRATION_PARAMETER_THROTTLE_TRIGGER_THRESHOLD),
+ params->throttle_trigger_threshold);
assert(params->has_cpu_throttle_initial);
monitor_printf(mon, "%s: %u\n",
MigrationParameter_str(MIGRATION_PARAMETER_CPU_THROTTLE_INITIAL),
@@ -1254,6 +1258,9 @@ void hmp_migrate_set_parameter(Monitor *mon, const QDict *qdict)
p->has_decompress_threads = true;
visit_type_int(v, param, &p->decompress_threads, &err);
break;
+ case MIGRATION_PARAMETER_THROTTLE_TRIGGER_THRESHOLD:
+ p->has_throttle_trigger_threshold = true;
+ visit_type_int(v, param, &p->throttle_trigger_threshold, &err);
case MIGRATION_PARAMETER_CPU_THROTTLE_INITIAL:
p->has_cpu_throttle_initial = true;
visit_type_int(v, param, &p->cpu_throttle_initial, &err);
diff --git a/qapi/migration.json b/qapi/migration.json
index d44d99cd78..0d1c0712ca 100644
--- a/qapi/migration.json
+++ b/qapi/migration.json
@@ -540,6 +540,10 @@
# compression, so set the decompress-threads to the number about 1/4
# of compress-threads is adequate.
#
+# @throttle-trigger-threshold: The ratio of bytes_dirty_period and bytes_xfer_period
+# to trigger throttling. It is expressed as percentage.
+# The default value is 50. (Since 5.0)
+#
# @cpu-throttle-initial: Initial percentage of time guest cpus are throttled
# when migration auto-converge is activated. The
# default value is 20. (Since 2.7)
@@ -625,7 +629,7 @@
'data': ['announce-initial', 'announce-max',
'announce-rounds', 'announce-step',
'compress-level', 'compress-threads', 'decompress-threads',
- 'compress-wait-thread',
+ 'compress-wait-thread', 'throttle-trigger-threshold',
'cpu-throttle-initial', 'cpu-throttle-increment',
'tls-creds', 'tls-hostname', 'tls-authz', 'max-bandwidth',
'downtime-limit', 'x-checkpoint-delay', 'block-incremental',
@@ -660,6 +664,10 @@
#
# @decompress-threads: decompression thread count
#
+# @throttle-trigger-threshold: The ratio of bytes_dirty_period and bytes_xfer_period
+# to trigger throttling. It is expressed as percentage.
+# The default value is 50. (Since 5.0)
+#
# @cpu-throttle-initial: Initial percentage of time guest cpus are
# throttled when migration auto-converge is activated.
# The default value is 20. (Since 2.7)
@@ -752,6 +760,7 @@
'*compress-threads': 'int',
'*compress-wait-thread': 'bool',
'*decompress-threads': 'int',
+ '*throttle-trigger-threshold': 'int',
'*cpu-throttle-initial': 'int',
'*cpu-throttle-increment': 'int',
'*tls-creds': 'StrOrNull',
@@ -813,6 +822,10 @@
#
# @decompress-threads: decompression thread count
#
+# @throttle-trigger-threshold: The ratio of bytes_dirty_period and bytes_xfer_period
+# to trigger throttling. It is expressed as percentage.
+# The default value is 50. (Since 5.0)
+#
# @cpu-throttle-initial: Initial percentage of time guest cpus are
# throttled when migration auto-converge is activated.
# (Since 2.7)
@@ -905,6 +918,7 @@
'*compress-threads': 'uint8',
'*compress-wait-thread': 'bool',
'*decompress-threads': 'uint8',
+ '*throttle-trigger-threshold': 'uint8',
'*cpu-throttle-initial': 'uint8',
'*cpu-throttle-increment': 'uint8',
'*tls-creds': 'str',
--
2.24.1
^ permalink raw reply related [flat|nested] 7+ messages in thread
* [PULL 3/5] COLO: Optimize memory back-up process
2020-03-13 10:30 [PULL 0/5] migration queue Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 1/5] configure: Improve zstd test Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 2/5] migration/throttle: Add throttle-trig-thres migration parameter Dr. David Alan Gilbert (git)
@ 2020-03-13 10:30 ` Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 4/5] ram/colo: only record bitmap of dirty pages in COLO stage Dr. David Alan Gilbert (git)
` (2 subsequent siblings)
5 siblings, 0 replies; 7+ messages in thread
From: Dr. David Alan Gilbert (git) @ 2020-03-13 10:30 UTC (permalink / raw)
To: qemu-devel, zhang.zhanghailiang, zhukeqian1, quintela
From: zhanghailiang <zhang.zhanghailiang@huawei.com>
This patch will reduce the downtime of VM for the initial process,
Previously, we copied all these memory in preparing stage of COLO
while we need to stop VM, which is a time-consuming process.
Here we optimize it by a trick, back-up every page while in migration
process while COLO is enabled, though it affects the speed of the
migration, but it obviously reduce the downtime of back-up all SVM'S
memory in COLO preparing stage.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Message-Id: <20200224065414.36524-5-zhang.zhanghailiang@huawei.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
minor typo fixes
---
migration/colo.c | 3 +++
migration/ram.c | 66 +++++++++++++++++++++++++++++++++++-------------
migration/ram.h | 1 +
3 files changed, 53 insertions(+), 17 deletions(-)
diff --git a/migration/colo.c b/migration/colo.c
index 93c5a452fb..44942c4e23 100644
--- a/migration/colo.c
+++ b/migration/colo.c
@@ -26,6 +26,7 @@
#include "qemu/main-loop.h"
#include "qemu/rcu.h"
#include "migration/failover.h"
+#include "migration/ram.h"
#ifdef CONFIG_REPLICATION
#include "replication.h"
#endif
@@ -845,6 +846,8 @@ void *colo_process_incoming_thread(void *opaque)
*/
qemu_file_set_blocking(mis->from_src_file, true);
+ colo_incoming_start_dirty_log();
+
bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
fb = qemu_fopen_channel_input(QIO_CHANNEL(bioc));
object_unref(OBJECT(bioc));
diff --git a/migration/ram.c b/migration/ram.c
index 02cfd76d19..31f4a9dc09 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -2994,7 +2994,6 @@ int colo_init_ram_cache(void)
}
return -errno;
}
- memcpy(block->colo_cache, block->host, block->used_length);
}
}
@@ -3008,19 +3007,36 @@ int colo_init_ram_cache(void)
RAMBLOCK_FOREACH_NOT_IGNORED(block) {
unsigned long pages = block->max_length >> TARGET_PAGE_BITS;
-
block->bmap = bitmap_new(pages);
- bitmap_set(block->bmap, 0, pages);
}
}
- ram_state = g_new0(RAMState, 1);
- ram_state->migration_dirty_pages = 0;
- qemu_mutex_init(&ram_state->bitmap_mutex);
- memory_global_dirty_log_start();
+ ram_state_init(&ram_state);
return 0;
}
+/* TODO: duplicated with ram_init_bitmaps */
+void colo_incoming_start_dirty_log(void)
+{
+ RAMBlock *block = NULL;
+ /* For memory_global_dirty_log_start below. */
+ qemu_mutex_lock_iothread();
+ qemu_mutex_lock_ramlist();
+
+ memory_global_dirty_log_sync();
+ WITH_RCU_READ_LOCK_GUARD() {
+ RAMBLOCK_FOREACH_NOT_IGNORED(block) {
+ ramblock_sync_dirty_bitmap(ram_state, block);
+ /* Discard this dirty bitmap record */
+ bitmap_zero(block->bmap, block->max_length >> TARGET_PAGE_BITS);
+ }
+ memory_global_dirty_log_start();
+ }
+ ram_state->migration_dirty_pages = 0;
+ qemu_mutex_unlock_ramlist();
+ qemu_mutex_unlock_iothread();
+}
+
/* It is need to hold the global lock to call this helper */
void colo_release_ram_cache(void)
{
@@ -3040,9 +3056,7 @@ void colo_release_ram_cache(void)
}
}
}
- qemu_mutex_destroy(&ram_state->bitmap_mutex);
- g_free(ram_state);
- ram_state = NULL;
+ ram_state_cleanup(&ram_state);
}
/**
@@ -3356,7 +3370,7 @@ static int ram_load_precopy(QEMUFile *f)
while (!ret && !(flags & RAM_SAVE_FLAG_EOS)) {
ram_addr_t addr, total_ram_bytes;
- void *host = NULL;
+ void *host = NULL, *host_bak = NULL;
uint8_t ch;
/*
@@ -3387,20 +3401,35 @@ static int ram_load_precopy(QEMUFile *f)
RAM_SAVE_FLAG_COMPRESS_PAGE | RAM_SAVE_FLAG_XBZRLE)) {
RAMBlock *block = ram_block_from_stream(f, flags);
+ host = host_from_ram_block_offset(block, addr);
/*
- * After going into COLO, we should load the Page into colo_cache.
+ * After going into COLO stage, we should not load the page
+ * into SVM's memory directly, we put them into colo_cache firstly.
+ * NOTE: We need to keep a copy of SVM's ram in colo_cache.
+ * Previously, we copied all these memory in preparing stage of COLO
+ * while we need to stop VM, which is a time-consuming process.
+ * Here we optimize it by a trick, back-up every page while in
+ * migration process while COLO is enabled, though it affects the
+ * speed of the migration, but it obviously reduce the downtime of
+ * back-up all SVM'S memory in COLO preparing stage.
*/
- if (migration_incoming_in_colo_state()) {
- host = colo_cache_from_block_offset(block, addr);
- } else {
- host = host_from_ram_block_offset(block, addr);
+ if (migration_incoming_colo_enabled()) {
+ if (migration_incoming_in_colo_state()) {
+ /* In COLO stage, put all pages into cache temporarily */
+ host = colo_cache_from_block_offset(block, addr);
+ } else {
+ /*
+ * In migration stage but before COLO stage,
+ * Put all pages into both cache and SVM's memory.
+ */
+ host_bak = colo_cache_from_block_offset(block, addr);
+ }
}
if (!host) {
error_report("Illegal RAM offset " RAM_ADDR_FMT, addr);
ret = -EINVAL;
break;
}
-
if (!migration_incoming_in_colo_state()) {
ramblock_recv_bitmap_set(block, host);
}
@@ -3514,6 +3543,9 @@ static int ram_load_precopy(QEMUFile *f)
if (!ret) {
ret = qemu_file_get_error(f);
}
+ if (!ret && host_bak) {
+ memcpy(host_bak, host, TARGET_PAGE_SIZE);
+ }
}
ret |= wait_for_decompress_done();
diff --git a/migration/ram.h b/migration/ram.h
index a553d40751..5ceaff7cb4 100644
--- a/migration/ram.h
+++ b/migration/ram.h
@@ -66,5 +66,6 @@ int ram_dirty_bitmap_reload(MigrationState *s, RAMBlock *rb);
/* ram cache */
int colo_init_ram_cache(void);
void colo_release_ram_cache(void);
+void colo_incoming_start_dirty_log(void);
#endif
--
2.24.1
^ permalink raw reply related [flat|nested] 7+ messages in thread
* [PULL 4/5] ram/colo: only record bitmap of dirty pages in COLO stage
2020-03-13 10:30 [PULL 0/5] migration queue Dr. David Alan Gilbert (git)
` (2 preceding siblings ...)
2020-03-13 10:30 ` [PULL 3/5] COLO: Optimize memory back-up process Dr. David Alan Gilbert (git)
@ 2020-03-13 10:30 ` Dr. David Alan Gilbert (git)
2020-03-13 10:30 ` [PULL 5/5] migration: recognize COLO as part of activating process Dr. David Alan Gilbert (git)
2020-03-13 14:48 ` [PULL 0/5] migration queue Peter Maydell
5 siblings, 0 replies; 7+ messages in thread
From: Dr. David Alan Gilbert (git) @ 2020-03-13 10:30 UTC (permalink / raw)
To: qemu-devel, zhang.zhanghailiang, zhukeqian1, quintela
From: zhanghailiang <zhang.zhanghailiang@huawei.com>
It is only need to record bitmap of dirty pages while goes
into COLO stage.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Message-Id: <20200224065414.36524-6-zhang.zhanghailiang@huawei.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
---
migration/ram.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/migration/ram.c b/migration/ram.c
index 31f4a9dc09..c12cfdbe26 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -2742,7 +2742,7 @@ static inline void *host_from_ram_block_offset(RAMBlock *block,
}
static inline void *colo_cache_from_block_offset(RAMBlock *block,
- ram_addr_t offset)
+ ram_addr_t offset, bool record_bitmap)
{
if (!offset_in_ramblock(block, offset)) {
return NULL;
@@ -2758,7 +2758,8 @@ static inline void *colo_cache_from_block_offset(RAMBlock *block,
* It help us to decide which pages in ram cache should be flushed
* into VM's RAM later.
*/
- if (!test_and_set_bit(offset >> TARGET_PAGE_BITS, block->bmap)) {
+ if (record_bitmap &&
+ !test_and_set_bit(offset >> TARGET_PAGE_BITS, block->bmap)) {
ram_state->migration_dirty_pages++;
}
return block->colo_cache + offset;
@@ -3416,13 +3417,13 @@ static int ram_load_precopy(QEMUFile *f)
if (migration_incoming_colo_enabled()) {
if (migration_incoming_in_colo_state()) {
/* In COLO stage, put all pages into cache temporarily */
- host = colo_cache_from_block_offset(block, addr);
+ host = colo_cache_from_block_offset(block, addr, true);
} else {
/*
* In migration stage but before COLO stage,
* Put all pages into both cache and SVM's memory.
*/
- host_bak = colo_cache_from_block_offset(block, addr);
+ host_bak = colo_cache_from_block_offset(block, addr, false);
}
}
if (!host) {
--
2.24.1
^ permalink raw reply related [flat|nested] 7+ messages in thread