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

alexey pushed a commit to branch branch-1.18.x
in repository https://gitbox.apache.org/repos/asf/kudu.git

commit eb09a2428b8cd89bc4108922510a5d17e536291c
Author: Alexey Serbin <[email protected]>
AuthorDate: Tue Oct 8 17:07:25 2024 -0700

    [dynamic_multi_master-test] reduce the amount of generated data
    
    When I ran all C++ tests via ctest on one node having about 25GiBytes of
    available disk space, it unexpectedly ran out of space.  The culprit was
    dynamic_multi_master-test: the scenarios from the suite keep their test
    workloads running while waiting for events that might take a long time
    to happen, especially on slow or very busy machines.
    
    This patch updates the configuration of the TestWorkload objects to use
    a single writer thread (the default is four) in the TestBasicAddition
    scenario of the ParameterizedAutoAddMasterTest test.  In addition,
    a bit of extra delay is now added between writing batches of test rows.
    
    Change-Id: Id0f0005b8f3384ad05eee6bf65e17a0e4ed289f1
    Reviewed-on: http://gerrit.cloudera.org:8080/21935
    Tested-by: Alexey Serbin <[email protected]>
    Reviewed-by: Marton Greber <[email protected]>
    Reviewed-by: Attila Bukor <[email protected]>
    (cherry picked from commit dbb6d5320ad60016355d0c63b6dae1650ba4e414)
    Reviewed-on: http://gerrit.cloudera.org:8080/22000
    Reviewed-by: Abhishek Chennaka <[email protected]>
---
 src/kudu/master/dynamic_multi_master-test.cc | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/src/kudu/master/dynamic_multi_master-test.cc 
b/src/kudu/master/dynamic_multi_master-test.cc
index 1479ef736..beb825789 100644
--- a/src/kudu/master/dynamic_multi_master-test.cc
+++ b/src/kudu/master/dynamic_multi_master-test.cc
@@ -1797,6 +1797,12 @@ class ParameterizedAutoAddMasterTest : public 
AutoAddMasterTest,
 TEST_P(ParameterizedAutoAddMasterTest, TestBasicAddition) {
   TestWorkload w(cluster_.get());
   w.set_num_replicas(1);
+  // Using one thread and injecting delays between write batches to reduce
+  // the amount of data accumulated under the tablet server's WAL and data
+  // directories since this scenario might run for quite a long time,
+  // especially on slow machines.
+  w.set_num_write_threads(1);
+  w.set_write_interval_millis(5);
   w.Setup();
   w.Start();
   int num_masters = args_.orig_num_masters;

Reply via email to