[
https://issues.apache.org/jira/browse/PHOENIX-15?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13904856#comment-13904856
]
James Taylor commented on PHOENIX-15:
-------------------------------------
Thanks for finding *another* bug. Given that now we only pass through the
disableWAL boolean if it changes, a simpler way to implement this would be in
the ConnectionQueryServicesImpl.addColumn() method and add a private flush
method. We can surface a flush method on the ConnectionQueryServices interface
later if/when we decide to surface a FLUSH command.
What do you think, [~lhofhansl]?
{code}
if (result.getMutationCode() == MutationCode.COLUMN_NOT_FOUND
&& Boolean.FALSE.equals(PDataType.BOOLEAN.toObject(
MetaDataUtil.getMutationKVByteValue(m,PhoenixDatabaseMetaData.DISABLE_WAL_BYTES))))
{
HTableInterface ht = flushTable(table.getPhysicalName().getBytes());
}
{code}
> Allow forcing a memstore flush for a table
> ------------------------------------------
>
> Key: PHOENIX-15
> URL: https://issues.apache.org/jira/browse/PHOENIX-15
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Lars Hofhansl
> Attachments: phoenix-15-v2.txt, phoenix-15-v3.txt, phoenix-15.txt
>
>
> This is useful with PHOENIX-14.
> I.e. one could do a "bulk load" without writing to the WAL, following by a
> memstore flush in order to guarantee the data is on disk.
> [[email protected]] suggested: ALTER SESSION FLUSH. And somehow we'd
> need to indicate which table to flush.
> This probably needs a bit more discussion.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
