[ 
https://issues.apache.org/jira/browse/KUDU-2475?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16518686#comment-16518686
 ] 

Dan Burkert commented on KUDU-2475:
-----------------------------------

This can be reproduced with a single master and injecting faults into the sys 
catalog:

 
{code:java}
TEST_F(MasterHmsTest, TestSysCatalogWriteFailure) {
  // Create some tables which will be used later.
  ASSERT_OK(CreateKuduTable("default", "a"));
  ASSERT_OK(CreateKuduTable("default", "b"));

  // Ensure writes to the sys catalog fail.
  ASSERT_OK(cluster_->SetFlag(cluster_->master(), 
"sys_catalog_fail_during_write", "1.0"));

  // Attempt to create a table, ensuring the attempt fails.
  ASSERT_FALSE(CreateKuduTable("default", "c").ok());
  // The check below succeeds, because of the rollback mechanism during table 
creation.
  CheckTableDoesNotExist("default", "c");

  // Rename table a.
  unique_ptr<KuduTableAlterer> 
table_alterer(client_->NewTableAlterer("default.a"));
  Status s = table_alterer->RenameTo("default.a2")->Alter();

  // The alter succeeds:
  ASSERT_OK(s);

  // But the table isn't altered in Kudu (the check below fails):
  NO_FATALS(CheckTable("default", "a2"));

  // Delete table b.
  s = client_->DeleteTable("default.b");

  // The delete succeeds:
  ASSERT_OK(s);

  // But the table still exists in Kudu (the check below fails):
  NO_FATALS(CheckTableDoesNotExist("default", "b"));
}
{code}

> HMS Catalog consistency with multi-master
> -----------------------------------------
>
>                 Key: KUDU-2475
>                 URL: https://issues.apache.org/jira/browse/KUDU-2475
>             Project: Kudu
>          Issue Type: Improvement
>          Components: hms, master
>            Reporter: Dan Burkert
>            Assignee: Dan Burkert
>            Priority: Major
>
> There are potential issues in the current iteration of the HMS integration 
> which may cause clients to recieve acks for ALTER TABLE RENAME / DROP TABLE 
> operations which don't succeed.  See [Adar's 
> comment|https://gerrit.cloudera.org/c/8313/27/src/kudu/master/hms_notification_log_listener.cc#234]
>  on the notification log listener patch for more info.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to