memory management of large byte[]

[bculberson]

when using the Pooled Client Manager, never releasing the buffer is really problematic. for example if we have 50 in the pool, each one will grow to the size of the buggest object sent through it and that buffer is never released. this puts a large memory footprint on long running apps, like web services/web sites, especially when having some large data objects and having a large pool of connections.

[mythz]

Cool, well it's a lot simpler and reduces the complexity so I'm happy with it.
Also I didn't know Socket's could take ArraySegment<byte> - so thx for the tip :)

BTW Congratz :)
ServiceStack/ServiceStack@091c46e

[desunit]

@bculberson The latest changes have broken the following tests:

RedisBenchmarkTests.Measure_pipeline_speedup
RedisClientTests.Can_GetTypeIdsSet
RedisClientTests.Can_StoreAll_RedisClient
RedisClientTests.Can_StoreAll_RedisTypedClient

Exception is: System.Net.Sockets.SocketException : Message too long

Can you review them and adjust code to eliminate that problem?

[bculberson]

I ran those tests locally and could not recreate the SocketException. Do
you know how I may be able to recreate this?

Thanks!
Brad

On Tue, Aug 21, 2012 at 12:11 AM, Sergey Bogdanov
notifications@github.comwrote:

The latest changes have broken the following tests:

RedisBenchmarkTests.Measure_pipeline_speedup
RedisClientTests.Can_GetTypeIdsSet

RedisClientTests.Can_StoreAll_RedisClient

RedisClientTests.Can_StoreAll_RedisTypedClient

Exception is: Test(s) failed. System.Net.Sockets.SocketException : Message
too long

Can you review them and adjust code to eliminate that problem?

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/79#issuecomment-7892470.

[desunit]

Yes, same here. Couldn't reproduce it locally - ServiceStack* TeamCity server returned:

Test(s) failed. System.Net.Sockets.SocketException : Message too long
  at System.Net.Sockets.Socket.Send (IList`1 buffers, SocketFlags socketFlags) [0x00000] in <filename unknown>:0 
  at ServiceStack.Redis.RedisNativeClient.FlushSendBuffer () [0x00000] in <filename unknown>:0 
  at ServiceStack.Redis.RedisAllPurposePipeline.Flush () [0x00000] in <filename unknown>:0 
  at ServiceStack.Redis.Tests.RedisBenchmarkTests.Measure_pipeline_speedup () [0x00000] in <filename unknown>:0 
  at (wrapper managed-to-native) System.Reflection.MonoMethod:InternalInvoke (System.Reflection.MonoMethod,object,object[],System.Exception&)
  at System.Reflection.MonoMethod.Invoke (System.Object obj, BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] pa

[mythz]

Yeah those tests work for me too.

@desunit did you run your tests individually on Mono/Linux ?

[desunit]

@mythz yup, easy to reproduce with:

# /opt/mono-2.11.3/bin/nunit-console -fixture=ServiceStack.Redis.Tests.RedisBenchmarkTests ServiceStack.Redis.Tests.dll  

Full output:

Test Case Failures:
1) ServiceStack.Redis.Tests.RedisBenchmarkTests.Measure_pipeline_speedup : System.Net.Sockets.SocketException : Message too long
at System.Net.Sockets.Socket.Send (IList`1 buffers, SocketFlags socketFlags) [0x00018] in /home/sstack/build/mono-2.11.3/mcs/class/System/System.Net.Sockets/Socket_2_1.cs:1919
at ServiceStack.Redis.RedisNativeClient.FlushSendBuffer () [0x00000] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/src/ServiceStack.Redis/RedisNativeClient_Utils.cs:236
at ServiceStack.Redis.RedisAllPurposePipeline.Flush () [0x00000] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/src/ServiceStack.Redis/Pipeline/RedisAllPurposePipeline.cs:36
at ServiceStack.Redis.Tests.RedisBenchmarkTests.Measure_pipeline_speedup () [0x00145] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/tests/ServiceStack.Redis.Tests/RedisBenchmarkTests.cs:43
at (wrapper managed-to-native) System.Reflection.MonoMethod:InternalInvoke (System.Reflection.MonoMethod,object,object[],System.Exception&)
at System.Reflection.MonoMethod.Invoke (System.Object obj, BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) [0x000d0] in /home/sstack/build/mono-2.11.3/mcs/class/corlib/System.Reflection/MonoMethod.cs:229

[bculberson]

ill take a look at it in mono, it looks like it is specific to that
runtime... my guess is its in the usage of the other socket.Send which is
now sending in an IList<ArraySegment>. i should be able to easily
work around that.

On Tue, Aug 21, 2012 at 9:59 AM, Sergey Bogdanov
notifications@github.comwrote:

@mythz https://github.com/mythz yup, easy to reproduce with:

/opt/mono-2.11.3/bin/nunit-console -fixture=ServiceStack.Redis.Tests.RedisBenchmarkTests ServiceStack.Redis.Tests.dll

Full output:

Test Case Failures:

1. ServiceStack.Redis.Tests.RedisBenchmarkTests.Measure_pipeline_speedup : System.Net.Sockets.SocketException : Message too long
   at System.Net.Sockets.Socket.Send (IList`1 buffers, SocketFlags socketFlags) [0x00018] in /home/sstack/build/mono-2.11.3/mcs/class/System/System.Net.Sockets/Socket_2_1.cs:1919
   at ServiceStack.Redis.RedisNativeClient.FlushSendBuffer () [0x00000] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/src/ServiceStack.Redis/RedisNativeClient_Utils.cs:236
   at ServiceStack.Redis.RedisAllPurposePipeline.Flush () [0x00000] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/src/ServiceStack.Redis/Pipeline/RedisAllPurposePipeline.cs:36
   at ServiceStack.Redis.Tests.RedisBenchmarkTests.Measure_pipeline_speedup () [0x00145] in /usr/local/teamcity/buildAgent/work/a129dfcd8b79320d/tests/ServiceStack.Redis.Tests/RedisBenchmarkTests.cs:43
   at (wrapper managed-to-native) System.Reflection.MonoMethod:InternalInvoke (System.Reflection.MonoMethod,object,object[],System.Exception&)
   at System.Reflection.MonoMethod.Invoke (System.Object obj, BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) [0x000d0] in /home/sstack/build/mono-2.11.3/mcs/class/corlib/System.Reflection/MonoMethod.cs:229

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/79#issuecomment-7905957.

[desunit]

Yup, it looks like we have Linux specific issue
http://stackoverflow.com/questions/4795787/socket-error90message-too-long

[bculberson]

checked in a fix for it not using the socket.Send of the ArraySegments.