[
https://issues.apache.org/jira/browse/PHOENIX-3000?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lars Hofhansl updated PHOENIX-3000:
-----------------------------------
Attachment: 3000.txt
Here's a simple fix. Copy the key, unless it occupies more than 10% of the
array handed to us. Or in other words, if we can save 90% of HEAP during
aggregate we'll take the hit of making the copy.
In some tests I ran that made the difference between finishing and simply
running out of memory.
Note that it is not quite as simple. If the block is cached in HBase anyway -
and we're not using BLOCK_ENCODING - we would not consume more memory since we
made a copy, but the block is on the heap anyway.
> Reduce memory consumption during DISTINCT aggregation
> -----------------------------------------------------
>
> Key: PHOENIX-3000
> URL: https://issues.apache.org/jira/browse/PHOENIX-3000
> Project: Phoenix
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Attachments: 3000.txt
>
>
> In {{DistinctValueWithCountServerAggregator.aggregate}} we hold on the ptr
> handed to us from HBase.
> Note that this pointer points into an HFile Block, and hence we hold onto the
> entire block for the duration of the aggregation.
> If the column has high cardinality we might attempt holding the entire table
> in memory in the extreme case.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
