This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new 28e4c6936ca [metric](cloud) add metrics for get tablets and rowsets
(#51320)
28e4c6936ca is described below
commit 28e4c6936ca1f298c6ca3b1e947209b3285ea32b
Author: TengJianPing <tengjianp...@selectdb.com>
AuthorDate: Sat May 31 18:05:56 2025 +0800
[metric](cloud) add metrics for get tablets and rowsets (#51320)
---
be/src/cloud/config.cpp | 1 +
be/src/cloud/config.h | 1 +
be/src/pipeline/exec/olap_scan_operator.cpp | 34 +++++++++++++++++++++++++++++
be/src/util/doris_metrics.cpp | 6 +++++
be/src/util/doris_metrics.h | 3 +++
be/test/util/doris_metrics_test.cpp | 11 ++++++++++
6 files changed, 56 insertions(+)
diff --git a/be/src/cloud/config.cpp b/be/src/cloud/config.cpp
index bc5c90e6e94..b0f80835598 100644
--- a/be/src/cloud/config.cpp
+++ b/be/src/cloud/config.cpp
@@ -40,6 +40,7 @@ DEFINE_Int64(tablet_cache_capacity, "100000");
DEFINE_Int64(tablet_cache_shards, "16");
DEFINE_mInt32(tablet_sync_interval_s, "1800");
DEFINE_mInt32(init_scanner_sync_rowsets_parallelism, "10");
+DEFINE_mInt32(sync_rowsets_slow_threshold_ms, "1000");
DEFINE_mInt64(min_compaction_failure_interval_ms, "5000");
DEFINE_mInt64(base_compaction_freeze_interval_s, "7200");
diff --git a/be/src/cloud/config.h b/be/src/cloud/config.h
index 9e724082c9f..f7b85231cbd 100644
--- a/be/src/cloud/config.h
+++ b/be/src/cloud/config.h
@@ -71,6 +71,7 @@ DECLARE_Int64(tablet_cache_shards);
DECLARE_mInt32(tablet_sync_interval_s);
// parallelism for scanner init where may issue RPCs to sync rowset meta from
MS
DECLARE_mInt32(init_scanner_sync_rowsets_parallelism);
+DECLARE_mInt32(sync_rowsets_slow_threshold_ms);
// Cloud compaction config
DECLARE_mInt64(min_compaction_failure_interval_ms);
diff --git a/be/src/pipeline/exec/olap_scan_operator.cpp
b/be/src/pipeline/exec/olap_scan_operator.cpp
index 58a7730e8e5..b642c857dd1 100644
--- a/be/src/pipeline/exec/olap_scan_operator.cpp
+++ b/be/src/pipeline/exec/olap_scan_operator.cpp
@@ -513,6 +513,40 @@ Status OlapScanLocalState::hold_tablets() {
COUNTER_UPDATE(_sync_rowset_get_remote_delete_bitmap_rpc_timer,
sync_stats.get_remote_delete_bitmap_rpc_ns);
}
+ auto time_ms = duration_ns / 1000 / 1000;
+ if (time_ms >= config::sync_rowsets_slow_threshold_ms) {
+
DorisMetrics::instance()->get_remote_tablet_slow_time_ms->increment(time_ms);
+ DorisMetrics::instance()->get_remote_tablet_slow_cnt->increment(1);
+ LOG_WARNING("get tablet takes too long")
+ .tag("query_id",
print_id(PipelineXLocalState<>::_state->query_id()))
+ .tag("node_id", _parent->node_id())
+ .tag("total_time", PrettyPrinter::print(duration_ns,
TUnit::TIME_NS))
+ .tag("num_tablets", _tablets.size())
+ .tag("tablet_meta_cache_hit",
_sync_rowset_tablet_meta_cache_hit->value())
+ .tag("tablet_meta_cache_miss",
_sync_rowset_tablet_meta_cache_miss->value())
+ .tag("get_remote_tablet_meta_rpc_time",
+ PrettyPrinter::print(
+
_sync_rowset_get_remote_tablet_meta_rpc_timer->value(),
+ TUnit::TIME_NS))
+ .tag("remote_rowsets_num",
_sync_rowset_get_remote_rowsets_num->value())
+ .tag("get_remote_rowsets_rpc_time",
+
PrettyPrinter::print(_sync_rowset_get_remote_rowsets_rpc_timer->value(),
+ TUnit::TIME_NS))
+ .tag("local_delete_bitmap_rowsets_num",
+
_sync_rowset_get_local_delete_bitmap_rowsets_num->value())
+ .tag("remote_delete_bitmap_rowsets_num",
+
_sync_rowset_get_remote_delete_bitmap_rowsets_num->value())
+ .tag("remote_delete_bitmap_key_count",
+
_sync_rowset_get_remote_delete_bitmap_key_count->value())
+ .tag("remote_delete_bitmap_bytes",
+
PrettyPrinter::print(_sync_rowset_get_remote_delete_bitmap_bytes->value(),
+ TUnit::BYTES))
+ .tag("get_remote_delete_bitmap_rpc_time",
+ PrettyPrinter::print(
+
_sync_rowset_get_remote_delete_bitmap_rpc_timer->value(),
+ TUnit::TIME_NS));
+ }
+
} else {
for (size_t i = 0; i < _scan_ranges.size(); i++) {
int64_t version = 0;
diff --git a/be/src/util/doris_metrics.cpp b/be/src/util/doris_metrics.cpp
index bed4624ba9c..653ec275ae9 100644
--- a/be/src/util/doris_metrics.cpp
+++ b/be/src/util/doris_metrics.cpp
@@ -214,6 +214,9 @@
DEFINE_COUNTER_METRIC_PROTOTYPE_2ARG(runtime_filter_consumer_wait_ready_ms,
MetricUnit::MILLISECONDS);
DEFINE_GAUGE_CORE_METRIC_PROTOTYPE_2ARG(runtime_filter_consumer_timeout_num,
MetricUnit::NOUNIT);
+DEFINE_COUNTER_METRIC_PROTOTYPE_2ARG(get_remote_tablet_slow_time_ms,
MetricUnit::MILLISECONDS);
+DEFINE_COUNTER_METRIC_PROTOTYPE_2ARG(get_remote_tablet_slow_cnt,
MetricUnit::NOUNIT);
+
const std::string DorisMetrics::_s_registry_name = "doris_be";
const std::string DorisMetrics::_s_hook_name = "doris_metrics";
@@ -353,6 +356,9 @@ DorisMetrics::DorisMetrics() :
_metric_registry(_s_registry_name) {
INT_GAUGE_METRIC_REGISTER(_server_metric_entity,
runtime_filter_consumer_ready_num);
INT_COUNTER_METRIC_REGISTER(_server_metric_entity,
runtime_filter_consumer_wait_ready_ms);
INT_GAUGE_METRIC_REGISTER(_server_metric_entity,
runtime_filter_consumer_timeout_num);
+
+ INT_COUNTER_METRIC_REGISTER(_server_metric_entity,
get_remote_tablet_slow_time_ms);
+ INT_COUNTER_METRIC_REGISTER(_server_metric_entity,
get_remote_tablet_slow_cnt);
}
void DorisMetrics::initialize(bool init_system_metrics, const
std::set<std::string>& disk_devices,
diff --git a/be/src/util/doris_metrics.h b/be/src/util/doris_metrics.h
index 5b8515e7e5f..2a1827d06a1 100644
--- a/be/src/util/doris_metrics.h
+++ b/be/src/util/doris_metrics.h
@@ -241,6 +241,9 @@ public:
IntCounter* runtime_filter_consumer_wait_ready_ms = nullptr;
IntGauge* runtime_filter_consumer_timeout_num = nullptr;
+ IntCounter* get_remote_tablet_slow_time_ms = nullptr;
+ IntCounter* get_remote_tablet_slow_cnt = nullptr;
+
static DorisMetrics* instance() {
static DorisMetrics instance;
return &instance;
diff --git a/be/test/util/doris_metrics_test.cpp
b/be/test/util/doris_metrics_test.cpp
index 6e9969b1210..588805e71b1 100644
--- a/be/test/util/doris_metrics_test.cpp
+++ b/be/test/util/doris_metrics_test.cpp
@@ -178,6 +178,17 @@ TEST_F(DorisMetricsTest, Normal) {
EXPECT_TRUE(metric != nullptr);
EXPECT_STREQ("40", metric->to_string().c_str());
}
+ {
+
DorisMetrics::instance()->get_remote_tablet_slow_time_ms->increment(1000);
+ auto* metric =
server_entity->get_metric("get_remote_tablet_slow_time_ms");
+ EXPECT_TRUE(metric != nullptr);
+ EXPECT_STREQ("1000", metric->to_string().c_str());
+
+ DorisMetrics::instance()->get_remote_tablet_slow_cnt->increment(10);
+ metric = server_entity->get_metric("get_remote_tablet_slow_cnt");
+ EXPECT_TRUE(metric != nullptr);
+ EXPECT_STREQ("10", metric->to_string().c_str());
+ }
}
} // namespace doris
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
