Vihang Karajgaonkar has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/14319 )

Change subject: IMPALA-8984: Fix race condition in creating Kudu table
......................................................................


Patch Set 13:

> > (1 comment)
 >
 > This bug happened when submit several 'create if not exists'
 > queries concurrently to create kudu managed table(these queries are
 > same, at lease same table name, without Kudu-HMS Integration). If
 > one query obtain 'metastoreDdlLock_', this means already create
 > table in kudu storage(KuduCatalogOpExecutor.createManagedTable),
 > then other query prepare to create table in kudu storage, but
 > already exists, thus throw an exception , and this caused table
 > droped in kudu storage which created by first query. And then the
 > first query continue create table in hms which already obtain
 > 'metastoreDdlLock_', and to the end. These queries all execute
 > successful, but table only exist in hms, not in kudu storage.
 > This bug is hard to reprodue due to relying on threads execute
 > path, so I mock this situation in a test case: First create a kudu
 > managed table, then deleted it from catalog, still reserved in
 > metastore. Then submit the same ddl request again, and the original
 > code would lead to table been deleted in kudu storage, and thus we
 > would got an incompleted table when invalidated table. But with
 > this patch fixed, we would got an normal kudu table.

Thanks for the explanation. I now understand this race condition better.


--
To view, visit http://gerrit.cloudera.org:8080/14319
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I1a4047bcdaa6b346765b96e8c36bb747a2b0091d
Gerrit-Change-Number: 14319
Gerrit-PatchSet: 13
Gerrit-Owner: wangsheng <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Quanlong Huang <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-Reviewer: Vihang Karajgaonkar <[email protected]>
Gerrit-Reviewer: wangsheng <[email protected]>
Gerrit-Comment-Date: Thu, 12 Dec 2019 18:41:49 +0000
Gerrit-HasComments: No

Reply via email to