You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Enis Soztutar (JIRA)" <ji...@apache.org> on 2017/06/06 23:27:18 UTC

[jira] [Commented] (HBASE-18061) [C++] Fix retry logic in multi-get calls

    [ https://issues.apache.org/jira/browse/HBASE-18061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16039859#comment-16039859 ] 

Enis Soztutar commented on HBASE-18061:
---------------------------------------

For this: 
{code}
+  // TODO We don't get an info on whether to retry an exception or not. Should have a method to
{code}
It is unfortunate that the wire protocol for this does not contain information about whether the exception should be retried or not. We are just handling that in the java client layer by relying on the exception hierarchy to catch DoNotRetryExceptions in onError(). 

I have actually added bunch of java class names, and specific logic in the patch for HBASE-17907 (look at exception.h). Let me compile a list of all exceptions and enhance the ExceptionUtils::ShouldRetry(). 

bq. +  // We are just appending stack trace as NameBytesPair.value contains the entire info
Looking at ProtobufUtil.java #toException(), it seems that the value is the description rather than the stack trace. But we can re-use the stack trace field for this. 

exception-utils.h -> Sorry, in HBASE-17907 patch, I had to do the same thing. So we have this already in exception.h file. When you rebase you can use it instead. 

How are you testing this? 



> [C++] Fix retry logic in multi-get calls
> ----------------------------------------
>
>                 Key: HBASE-18061
>                 URL: https://issues.apache.org/jira/browse/HBASE-18061
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Enis Soztutar
>            Assignee: Sudeep Sunthankar
>             Fix For: HBASE-14850
>
>         Attachments: HBASE-18061.HBASE-14850.v1.patch
>
>
> HBASE-17576 adds multi-gets. There are a couple of todos to fix in the retry logic, and some unit testing to be done for the multi-gets. We'll do these in this issue. 



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