[
https://issues.apache.org/jira/browse/AVRO-3802?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
amy myo updated AVRO-3802:
--------------------------
Description:
Hi,
We are using apache.avro C# library for compression and decompression of large
avro files. Each file is 2-3MB and decompression of 20 files is using 2.1GB
memory and memory keep increasing around 200-300MB for follow up calls.
Attached is the memory profiler of three calls. Each call decompress the same
20 files. First call uses 2.1GB , increase to 2.37GB and then 2.68GB.
Call tree is showing most of the memory are used by `DeflatCodec.CopyTo`
method.
# `CopyTo` method is not using Array pool. Default Microsoft's stream `copyTo`
method is much more efficient.
# `inStream` and `outStream` memory streams are not in using statement.
# On `outstream` object creation, object can be initialized with defined
length/stream capacity.
was:
Hi,
We are using apache.avro C# library for compression and decompression of large
avro files. Each file has 2-3MB and decompression of 20 files is using 2.1GB
memory and memory keep increasing around 200-300MB for follow up calls.
Attached is the memory profiler of three calls. Each call decompressing the
same 20 files. First call uses 2.1GB , increase to 2.37GB and then 2.68GB.
The `decompression` method of the deflat
> Memory leak on decompressing deflate codec avro files
> -----------------------------------------------------
>
> Key: AVRO-3802
> URL: https://issues.apache.org/jira/browse/AVRO-3802
> Project: Apache Avro
> Issue Type: Bug
> Components: csharp
> Affects Versions: 1.11.1, 1.12.0, 1.11.2, 1.11.3
> Reporter: amy myo
> Priority: Major
> Attachments: call tree.PNG, deflate decompression memory profiler.png
>
>
> Hi,
> We are using apache.avro C# library for compression and decompression of
> large avro files. Each file is 2-3MB and decompression of 20 files is using
> 2.1GB memory and memory keep increasing around 200-300MB for follow up calls.
> Attached is the memory profiler of three calls. Each call decompress the same
> 20 files. First call uses 2.1GB , increase to 2.37GB and then 2.68GB.
> Call tree is showing most of the memory are used by `DeflatCodec.CopyTo`
> method.
> # `CopyTo` method is not using Array pool. Default Microsoft's stream
> `copyTo` method is much more efficient.
> # `inStream` and `outStream` memory streams are not in using statement.
> # On `outstream` object creation, object can be initialized with defined
> length/stream capacity.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
