1

be/src/olap/compaction.cpp

@@ -1127,6 +1127,18 @@ Status CloudCompactionMixin::execute_compact_impl(int64_t permits) {
 
     RETURN_IF_ERROR(merge_input_rowsets());
 
+    DBUG_EXECUTE_IF("CloudFullCompaction::modify_rowsets.wrong_rowset_id", {
+        DCHECK(compaction_type() == ReaderType::READER_FULL_COMPACTION);
+        RowsetId id;
+        id.version = 2;
+        id.hi = _output_rowset->rowset_meta()->rowset_id().hi + ((int64_t)(1) << 56);
+        id.mi = _output_rowset->rowset_meta()->rowset_id().mi;
+        id.lo = _output_rowset->rowset_meta()->rowset_id().lo;
+        _output_rowset->rowset_meta()->set_rowset_id(id);
+        LOG(INFO) << "[Debug wrong rowset id]:"
+                  << _output_rowset->rowset_meta()->rowset_id().to_string();
+    })
+
     RETURN_IF_ERROR(_engine.meta_mgr().commit_rowset(*_output_rowset->rowset_meta().get()));
 
     // 4. modify rowsets in memory

cloud/src/common/bvars.cpp

@@ -193,3 +193,6 @@ BvarStatusWithTag<long> g_bvar_checker_check_cost_s("checker", "check_cost_secon
 BvarStatusWithTag<long> g_bvar_checker_enqueue_cost_s("checker", "enqueue_cost_seconds");
 BvarStatusWithTag<long> g_bvar_checker_last_success_time_ms("checker", "last_success_time_ms");
 BvarStatusWithTag<long> g_bvar_checker_instance_volume("checker", "instance_volume");
+BvarStatusWithTag<long> g_bvar_inverted_checker_num_scanned("checker", "num_inverted_scanned");
+BvarStatusWithTag<long> g_bvar_inverted_checker_num_check_failed("checker",
+                                                                 "num_inverted_check_failed");

cloud/src/common/bvars.h

@@ -247,3 +247,5 @@ extern BvarStatusWithTag<long> g_bvar_checker_check_cost_s;
 extern BvarStatusWithTag<long> g_bvar_checker_enqueue_cost_s;
 extern BvarStatusWithTag<long> g_bvar_checker_last_success_time_ms;
 extern BvarStatusWithTag<long> g_bvar_checker_instance_volume;
+extern BvarStatusWithTag<long> g_bvar_inverted_checker_num_scanned;
+extern BvarStatusWithTag<long> g_bvar_inverted_checker_num_check_failed;

cloud/src/recycler/checker.cpp

@@ -168,15 +168,17 @@ int Checker::start() {
             auto ctime_ms =
                     duration_cast<milliseconds>(system_clock::now().time_since_epoch()).count();
             g_bvar_checker_enqueue_cost_s.put(instance_id, ctime_ms / 1000 - enqueue_time_s);
-            ret = checker->do_check();
+            int ret1 = checker->do_check();
+            if (ret1 == -1) return;
+            int ret2;
             if (config::enable_inverted_check) {
-                if (checker->do_inverted_check() != 0) ret = -1;
+                ret2 = checker->do_inverted_check();
+                if (ret2 == -1) return;
             }
-            if (ret == -1) return;
             // If instance checker has been aborted, don't finish this job
             if (!checker->stopped()) {
                 finish_instance_recycle_job(txn_kv_.get(), check_job_key, instance.instance_id(),
-                                            ip_port_, ret == 0, ctime_ms);
+                                            ip_port_, ret1 == 0 && ret2 == 0, ctime_ms);
             }
             {
                 std::lock_guard lock(mtx_);
@@ -606,6 +608,8 @@ int InstanceChecker::do_inverted_check() {
     using namespace std::chrono;
     auto start_time = steady_clock::now();
     std::unique_ptr<int, std::function<void(int*)>> defer_log_statistics((int*)0x01, [&](int*) {
+        g_bvar_inverted_checker_num_scanned.put(instance_id_, num_scanned);
+        g_bvar_inverted_checker_num_check_failed.put(instance_id_, num_check_failed);
         auto cost = duration<float>(steady_clock::now() - start_time).count();
         LOG(INFO) << "inverted check instance objects finished, cost=" << cost
                   << "s. instance_id=" << instance_id_ << " num_scanned=" << num_scanned
@@ -719,7 +723,7 @@ int InstanceChecker::do_inverted_check() {
             return -1;
         }
     }
-    return num_check_failed == 0 ? 0 : -1;
+    return num_check_failed == 0 ? 0 : -2;
 }
 
 } // namespace doris::cloud