[
https://issues.apache.org/jira/browse/HIVE-21509?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16801716#comment-16801716
]
Adam Szita commented on HIVE-21509:
-----------------------------------
I've attached a proposed solution (work in progress) that would fix this
problem, see [^HIVE-21509.0.wip.patch].
It adds a reference counter to ColumnVectors which is increased before
beginning to write and decreased once the CV has been passed to the actual
writer. EncodedDataConsumer will check this counter and will not return CVs
into the object pool whose ref count is greater than zero.
> LLAP may cache corrupted column vectors and return wrong query result
> ---------------------------------------------------------------------
>
> Key: HIVE-21509
> URL: https://issues.apache.org/jira/browse/HIVE-21509
> Project: Hive
> Issue Type: Bug
> Components: llap
> Reporter: Adam Szita
> Assignee: Adam Szita
> Priority: Major
> Attachments: HIVE-21509.0.wip.patch
>
>
> In some scenarios, LLAP might store column vectors in cache that are getting
> reused and reset just before their original content would be written.
> The issue is a concurrency issue and is thereby flaky. It is not easy to
> reproduce, but the odds of surfacing this issue can by improved by setting
> LLAP executor and IO thread counts this way:
> * set hive.llap.daemon.num.executors=32;
> * set hive.llap.io.threadpool.size=1;
> * using TPCDS input data of store_sales table, have at least a couple of
> 100k's of rows, and use text format:
> {code:java}
> ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe'
> WITH SERDEPROPERTIES ( 'field.delim'='|', 'serialization.format'='|')
> STORED AS INPUTFORMAT 'org.apache.hadoop.mapred.TextInputFormat'
> OUTPUTFORMAT
> 'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'{code}
> * having more splits increases the issue showing itself, so it is worth to
> _set tez.grouping.min-size=1024; set tez.grouping.max-size=1024;_
> * run query on this this table: select min(ss_sold_date_sk) from store_sales;
> The first query result is correct (2450816 in my case). Repeating the query
> will trigger reading from LLAP cache and produce a wrong result: 0.
> If one wants to make sure of running into this issue, place a
> Thread.sleep(250) at the beginning of VectorDeserializeOrcWriter#run().
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
