[
https://issues.apache.org/jira/browse/HIVE-18662?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eugene Koifman updated HIVE-18662:
----------------------------------
Description:
OrcRecordUpdater.KeyIndexBuilder stores an index in ORC footer where each entry
is the last ROW__ID of each stripe. In acid1 this is used to filter the events
from delta file when merging with part of the base.
as can be seen in {{TestTxnCommands.testVersioning()}} (added in HIVE-18659)
the {{hive.acid.key.index}} is empty.
This is because very little data is written and WriterImpl.flushStripe() is not
called except when {
{WriterImpl.close()}
is called. In the later, {{WriterCallback.preFooterWrite()}} is called before
{{preStripeWrite}} and so KeyIndexBuilder.preFooterWriter() records nothing in
\{{hive.acid.key.index}}
need to investigate if this is an issue, in particular acid 2
was:
OrcRecordUpdater.KeyIndexBuilder stores an index in ORC footer where each entry
is the last ROW__ID of each stripe. In acid1 this is used to filter the events
from delta file when merging with part of the base.
as can be seen in \{{TestTxnCommands.testVersioning()}} (added in HIVE-18659)
the \{{hive.acid.key.index}} is empty.
This is because very little data is written and WriterImpl.flushStripe() is not
called except when \{{WriterImpl.close()} is called. In the later,
\{{WriterCallback.preFooterWrite()}} is called before \{{preStripeWrite}} and
so KeyIndexBuilder.preFooterWriter() records nothing in \{{hive.acid.key.index}}
> hive.acid.key.index is missing entries
> --------------------------------------
>
> Key: HIVE-18662
> URL: https://issues.apache.org/jira/browse/HIVE-18662
> Project: Hive
> Issue Type: Bug
> Components: Transactions
> Reporter: Eugene Koifman
> Priority: Major
>
> OrcRecordUpdater.KeyIndexBuilder stores an index in ORC footer where each
> entry is the last ROW__ID of each stripe. In acid1 this is used to filter
> the events from delta file when merging with part of the base.
>
> as can be seen in {{TestTxnCommands.testVersioning()}} (added in HIVE-18659)
> the {{hive.acid.key.index}} is empty.
>
> This is because very little data is written and WriterImpl.flushStripe() is
> not called except when {
> {WriterImpl.close()}
> is called. In the later, {{WriterCallback.preFooterWrite()}} is called
> before {{preStripeWrite}} and so KeyIndexBuilder.preFooterWriter() records
> nothing in \{{hive.acid.key.index}}
>
> need to investigate if this is an issue, in particular acid 2
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
