[kudu] 02/03: [www] add slow scans show control

[la...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

laiyingchun pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/kudu.git

commit 97ec58742182bb8175dbf38a94027f845e806926
Author: kedeng <kd...@gmail.com>
AuthorDate: Tue Feb 7 17:07:49 2023 +0800

    [www] add slow scans show control
    
    The slow scans show functionality may lead to full scanning and affecting
    normal Kudu service unexpectedly. Add --show_slow_scans flag to control
    whether to show slow scans on the /scans page of web and record the
    slow scans in the log.
    
    The default value of the flag --show_slow_scans is false, which means
    slow scans show is disabled by default.
    
    Change-Id: Ia96f80561a4c889cbdd1c6dc992184981be86fb6
    Reviewed-on: http://gerrit.cloudera.org:8080/19480
    Reviewed-by: Yingchun Lai <la...@apache.org>
    Tested-by: Kudu Jenkins
---
 src/kudu/client/scan_token-test.cc | 21 +++++++++++++++++++++
 src/kudu/tserver/scanners.cc       | 24 +++++++++++++++++++-----
 2 files changed, 40 insertions(+), 5 deletions(-)

diff --git a/src/kudu/client/scan_token-test.cc b/src/kudu/client/scan_token-test.cc
index 165190fe7..0caba95ec 100644
--- a/src/kudu/client/scan_token-test.cc
+++ b/src/kudu/client/scan_token-test.cc
@@ -75,6 +75,7 @@
 DECLARE_bool(tserver_enforce_access_control);
 DECLARE_int32(scanner_inject_latency_on_each_batch_ms);
 DECLARE_int32(slow_scanner_threshold_ms);
+DECLARE_bool(show_slow_scans);
 
 METRIC_DECLARE_histogram(handler_latency_kudu_master_MasterService_GetTableSchema);
 METRIC_DECLARE_histogram(handler_latency_kudu_master_MasterService_GetTableLocations);
@@ -367,6 +368,10 @@ class ScanTokenTest : public KuduTest {
 };
 
 TEST_F(ScanTokenTest, SlowScansListTest) {
+  // Slow scans show is disabled by default.
+  ASSERT_FALSE(FLAGS_show_slow_scans);
+  FLAGS_show_slow_scans = true;
+
   constexpr const char* const kTableName = "slow_scans_show";
   // Create schema
   KuduSchema schema;
@@ -418,6 +423,22 @@ TEST_F(ScanTokenTest, SlowScansListTest) {
     ASSERT_EQ(2, GetCompletedScansCount());
     ASSERT_EQ(1, GetSlowScansCount());
   }
+
+  {
+    // Disable the slow scans show.
+    FLAGS_show_slow_scans = false;
+    vector<KuduScanToken*> tokens;
+    ElementDeleter deleter(&tokens);
+    ASSERT_OK(KuduScanTokenBuilder(table.get()).Build(&tokens));
+
+    // Create a slow scan scenarios.
+    FLAGS_scanner_inject_latency_on_each_batch_ms = 50;
+    FLAGS_slow_scanner_threshold_ms = 40;
+    ASSERT_EQ(200, CountRows(tokens));
+    ASSERT_EQ(3, GetCompletedScansCount());
+    // If disable the slow scans, we can only get 0, which represents the count of slow scans.
+    ASSERT_EQ(0, GetSlowScansCount());
+  }
 }
 
 TEST_F(ScanTokenTest, TestScanTokens) {
diff --git a/src/kudu/tserver/scanners.cc b/src/kudu/tserver/scanners.cc
index d68aa161b..d28f1a4aa 100644
--- a/src/kudu/tserver/scanners.cc
+++ b/src/kudu/tserver/scanners.cc
@@ -65,8 +65,13 @@ DEFINE_int32(completed_scan_history_count, 10,
              "latest scans will be shown on the tablet server's scans dashboard.");
 TAG_FLAG(completed_scan_history_count, experimental);
 
-// TODO(kedeng) : Add flag to control the display of slow scans, and avoid full scanning
-//                affecting normal Kudu service without perception.
+DEFINE_bool(show_slow_scans, false,
+            "Whether to show slow scans on the /scans page of web or record it in the log. "
+            "Please note that once set to true, full table scans may occur, which may affect "
+            "the normal Kudu service unexpectedly.");
+TAG_FLAG(show_slow_scans, advanced);
+TAG_FLAG(show_slow_scans, runtime);
+
 DEFINE_int32(slow_scanner_threshold_ms, 60 * 1000L, // 1 minute
              "Number of milliseconds for the threshold of slow scan.");
 TAG_FLAG(slow_scanner_threshold_ms, advanced);
@@ -171,7 +176,12 @@ void ScannerManager::RunCollectAndRemovalThread() {
       }
       shutdown_cv_.WaitFor(MonoDelta::FromMicroseconds(FLAGS_scanner_gc_check_interval_us));
     }
-    CollectSlowScanners();
+
+    if (FLAGS_show_slow_scans) {
+      // Control the collection of slow scans to avoid full scanning affecting normal Kudu
+      // service without perception.
+      CollectSlowScanners();
+    }
     RemoveExpiredScanners();
   }
 }
@@ -315,6 +325,12 @@ vector<SharedScanDescriptor> ScannerManager::ListScans() const {
 }
 
 vector<SharedScanDescriptor> ScannerManager::ListSlowScans() const {
+  vector<SharedScanDescriptor> ret;
+  if (!FLAGS_show_slow_scans) {
+    LOG(INFO) << "Slow scans show is disabled. Set --show_slow_scans to enable it.";
+    return ret;
+  }
+
   // Get all the scans first.
   unordered_map<string, SharedScanDescriptor> scans;
   {
@@ -324,7 +340,6 @@ vector<SharedScanDescriptor> ScannerManager::ListSlowScans() const {
     }
   }
 
-  vector<SharedScanDescriptor> ret;
   ret.reserve(scans.size());
   AppendValuesFromMap(scans, &ret);
 
@@ -357,7 +372,6 @@ void ScannerManager::CollectSlowScanners() {
 
       MonoDelta delta_time = now - start_time -
           MonoDelta::FromMilliseconds(slow_scanner_threshold);
-      // TODO(kedeng) : Add flag to control whether to print this log.
       LOG(INFO) << Substitute(
           "Slow scanner id: $0, of tablet $1, "
           "exceed the time threshold $2 ms for $3 ms.",