phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-4028) Provide option to not throw index write failure back to client
Date Wed, 19 Jul 2017 00:46:00 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-4028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16092418#comment-16092418
] 

James Taylor commented on PHOENIX-4028:
---------------------------------------

No need to add a new attribute to mutations. Just pass through IndexMetaData to IndexFailurePolicy.handleFailure
method and check ignoreNewerMutations (which means isPartialRebuild - feel free to rename
it) here:
{code}
       } finally {
           if (!throwing) {
           	IOException ioException = ServerUtil.wrapInDoNotRetryIOException("Unable to update
the following indexes: " + attempted.keySet(), cause, timestamp);
           	if (throwIndexWriteFailure || indexMetaData.ignoreNewerMutations()) { // Always
throw write failure if rebuilding index
{code}

> Provide option to not throw index write failure back to client
> --------------------------------------------------------------
>
>                 Key: PHOENIX-4028
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-4028
>             Project: Phoenix
>          Issue Type: Improvement
>            Reporter: James Taylor
>            Assignee: James Taylor
>             Fix For: 4.12.0
>
>         Attachments: PHOENIX-4028.patch, PHOENIX-4028_v2.patch, PHOENIX-4028_v3.patch,
PHOENIX-4028_v4.patch, PHOENIX-4028_v5_wip.patch, PHOENIX-4028_wip.patch
>
>
> Much like our DISABLE_INDEX_ON_WRITE_FAILURE and REBUILD_INDEX_ON_WRITE_FAILURE table
properties, we need a THROW_INDEX_WRITE_FAILURE boolean option that can be used to prevent
the index write from being thrown back to the client. In this case, the index failure policy
would still be executed (i.e. disabling the index on a write failure), but any retry logic
for the client would be avoided. The index would be eventually consistent based on the background
partial index rebuild thread.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message