[jira] [Commented] (HBASE-18178) [C++] Retrying meta location lookup and zookeeper connection

["Ted Yu (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-18178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16044891#comment-16044891 ] 

Ted Yu commented on HBASE-18178:
--------------------------------

{code}
+  // TODO: RegionLocation should propagate through these method chains as a shared_ptr.
+  // Otherwise, it may get deleted underneat us. We are just copying for now.
{code}
Is the above going to be done in next patch or new JIRA ?
{code}
+  static constexpr const uint32_t kDefaultClientRetriesNumber = 35;
{code}
Why is the retry number increased ?
{code}
+    LOG(INFO) << "Closed connection to ZooKeeper.";
{code}
Change the log level. (There're other LOG(INFO)'s in the patch)
{code}
+        throw ew;
+        return static_cast<std::unique_ptr<Response>>(nullptr);
{code}
Is the return statement effective ?





> [C++] Retrying meta location lookup and zookeeper connection 
> -------------------------------------------------------------
>
>                 Key: HBASE-18178
>                 URL: https://issues.apache.org/jira/browse/HBASE-18178
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>         Attachments: hbase-18178-v1.patch
>
>
> Currently location-cache can only do a single lookup to meta. If meta location changes or we have zookeeper connection problems, we never retry. 



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)