[improve](cloud-mow)Add delete bitmap metrics for MS (#47047)

When reading or writing delete bitmap fail, record failed count to metrics.
apache · Jan 26, 2025 · 92e2285 · 92e2285
1 parent c066d4f
commit 92e2285
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 0 deletions.
diff --git a/cloud/src/common/bvars.cpp b/cloud/src/common/bvars.cpp
@@ -85,6 +85,11 @@ BvarLatencyRecorderWithTag g_bvar_ms_finish_tablet_job("ms", "finish_tablet_job"
 BvarLatencyRecorderWithTag g_bvar_ms_get_cluster_status("ms", "get_cluster_status");
 BvarLatencyRecorderWithTag g_bvar_ms_set_cluster_status("ms", "set_cluster_status");
 BvarLatencyRecorderWithTag g_bvar_ms_check_kv("ms", "check_kv");
+bvar::Adder<int64_t> g_bvar_update_delete_bitmap_fail_counter;
+bvar::Window<bvar::Adder<int64_t> > g_bvar_update_delete_bitmap_fail_counter_minute("ms", "update_delete_bitmap_fail", &g_bvar_update_delete_bitmap_fail_counter, 60);
+bvar::Adder<int64_t> g_bvar_get_delete_bitmap_fail_counter;
+bvar::Window<bvar::Adder<int64_t> > g_bvar_get_delete_bitmap_fail_counter_minute("ms", "get_delete_bitmap_fail", &g_bvar_get_delete_bitmap_fail_counter, 60);
+
 
 // recycler's bvars
 // TODO: use mbvar for per instance, https://github.com/apache/brpc/blob/master/docs/cn/mbvar_c++.md

diff --git a/cloud/src/common/bvars.h b/cloud/src/common/bvars.h
@@ -160,6 +160,8 @@ extern BvarLatencyRecorderWithTag g_bvar_ms_get_rl_task_commit_attach;
 extern BvarLatencyRecorderWithTag g_bvar_ms_reset_rl_progress;
 extern BvarLatencyRecorderWithTag g_bvar_ms_get_txn_id;
 extern BvarLatencyRecorderWithTag g_bvar_ms_check_kv;
+extern bvar::Adder<int64_t> g_bvar_update_delete_bitmap_fail_counter;
+extern bvar::Adder<int64_t> g_bvar_get_delete_bitmap_fail_counter;
 
 // recycler's bvars
 extern BvarStatusWithTag<int64_t> g_bvar_recycler_recycle_index_earlest_ts;

diff --git a/cloud/src/meta-service/meta_service.cpp b/cloud/src/meta-service/meta_service.cpp
@@ -1921,6 +1921,7 @@ void MetaServiceImpl::update_delete_bitmap(google::protobuf::RpcController* cont
                    << " put_size=" << txn->put_bytes() << " num_put_keys=" << txn->num_put_keys()
                    << " txn_size=" << txn->approximate_bytes();
                 msg = ss.str();
+                g_bvar_update_delete_bitmap_fail_counter << 1;
                 return;
             }
             current_key_count = 0;
@@ -1963,6 +1964,7 @@ void MetaServiceImpl::update_delete_bitmap(google::protobuf::RpcController* cont
            << " delete_bitmap_value=" << current_value_count << " put_size=" << txn->put_bytes()
            << " num_put_keys=" << txn->num_put_keys() << " txn_size=" << txn->approximate_bytes();
         msg = ss.str();
+        g_bvar_update_delete_bitmap_fail_counter << 1;
         return;
     }
     LOG(INFO) << "update_delete_bitmap tablet_id=" << tablet_id << " lock_id=" << request->lock_id()
@@ -2075,6 +2077,7 @@ void MetaServiceImpl::get_delete_bitmap(google::protobuf::RpcController* control
                 ss << "internal error, failed to get delete bitmap, internal round=" << round
                    << ", ret=" << err;
                 msg = ss.str();
+                g_bvar_get_delete_bitmap_fail_counter << 1;
                 return;
             }
 
@@ -2118,6 +2121,7 @@ void MetaServiceImpl::get_delete_bitmap(google::protobuf::RpcController* control
                    << ",exceed max byte";
                 msg = ss.str();
                 LOG(WARNING) << msg;
+                g_bvar_get_delete_bitmap_fail_counter << 1;
                 return;
             }
             round++;