You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
[#27420] docdb: Fix flaky test MasterPathHandlersItest.TestClusterBalancerWarnings
Summary:
This one is the same issue as D44361. After D43208, the advisory lock table and transactions table are now created by CatalogManagerBgTasks. In slower builds (e.g., TSAN or ASAN), these tables might be created while only 2 out of 3 tservers are live. This can lead to them initially being created with only 2 replicas, making them under-replicated. The Load Balancer will then later handle the under replicated tablet.
Thus, we need more wait time before checking the Warnings Summary table to account for the time required by the Load Balancer to process the under-replicated advisory lock and transaction tablets.
The test was previously flaky due to insufficient wait time, as seen in the failure below:
```
/share/jenkins/workspace/github-yugabyte-db-alma8-master-clang19-tsan/yugabyte-db/src/yb/integration-tests/master_path_handlers-itest.cc:1751
Expected equality of these values:
rows.size()
Which is: 0
1
```
Jira: DB-16959
Test Plan: ./yb_build.sh asan --cxx-test integration-tests_master_path_handlers-itest --gtest_filter MasterPathHandlersItest.TestClusterBalancerWarnings
Reviewers: asrivastava
Reviewed By: asrivastava
Subscribers: ybase
Differential Revision: https://phorge.dev.yugabyte.com/D44377
0 commit comments