[jira] [Commented] (KUDU-2884) 'kudu hms fix' CLI tool mistakenly reports on non-matching master addresses

["ASF subversion and git services (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KUDU-2884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17193166#comment-17193166 ] 

ASF subversion and git services commented on KUDU-2884:
-------------------------------------------------------

Commit bd966e2b99a2b5f2a9752f61c8fe84a56d5d2307 in kudu's branch refs/heads/master from Grant Henke
[ https://gitbox.apache.org/repos/asf?p=kudu.git;h=bd966e2 ]

KUDU-2884: [hms] Improve master address matching

This patch improves the master address matching in the
`kudu hms fix` tool. Instead of expecting a literal match
of the address strings, it expects the match to be functionally
the same. This means that the addresses can be missing the
port (using the default), duplicated, and re-ordered. The
HMS will still consider them the same if they match the
clusters masters functionally.

Though the default port is considered when matching master
addresses, it is not included in the automated hms tool tests
given the tests use random ephemeral ports. Instead a test
for MasterAddressesToSet and UnorderedHostPortSet equality
was added.

Change-Id: I00059ff1b5134f7e5d6ea8be4fdf0701aa56b0c7
Reviewed-on: http://gerrit.cloudera.org:8080/16420
Reviewed-by: Andrew Wong <aw...@cloudera.com>
Reviewed-by: Alexey Serbin <as...@cloudera.com>
Tested-by: Kudu Jenkins


> 'kudu hms fix' CLI tool mistakenly reports on non-matching master addresses
> ---------------------------------------------------------------------------
>
>                 Key: KUDU-2884
>                 URL: https://issues.apache.org/jira/browse/KUDU-2884
>             Project: Kudu
>          Issue Type: Bug
>          Components: CLI, hms
>            Reporter: Alexey Serbin
>            Priority: Major
>
> When running the {{kudu hms fix}} against a Kudu cluster with managed Kudu tables created via {{impala-shell}}, the tool misinterprets master addresses being non-equal since from one side it gets the addresses with the default port number from CLI context, and from the other side the default port number is omitted (HMS metainfo).  As a result, the tool output an error and fails.
> {noformat}
> sudo -u kudu kudu hms fix a.b.c.com --drop_orphan_hms_tables --ignore_other_clusters=false
> I0630 17:29:55.427356  3689 tool_action_hms.cc:602] Skipping drop of orphan HMS table db_name.table_name with master addresses parameter a.b.c.com because it does not match the --master_addresses argument (use --force to skip this check)
> {noformat}
> Running the tool with {{--force}} might might introduce some issues if there are some table which indeed have different set of master addresses recorded in their HMS metadata.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)